Content uploaded by Daniel J Povinelli
Author content
All content in this area was uploaded by Daniel J Povinelli on Jul 15, 2019
Content may be subject to copyright.
INFANCY, 11(2), 157–174
Copyright © 2007, Lawrence Erlbaum Associates, Inc.
The Development of Body
Self-Awareness
Chris Moore, Jennifer Mealiea,
and Nancy Garon
Department of Psychology
Dalhousie University
Daniel J. Povinelli
Department of Cognitive Evolution Group
University of Louisiana, Lafayette
Two experiments examined toddlers’ performance on a new task designed to examine
the development of body self-awareness. The new task was conceived from observations
by Piaget (1953/1977) and theoretical work from Povinelli and Cant (1995) and
involved a toy shopping cart to the back of which a small mat had been attached.
Children were asked to push the cart to their mothers but in attempting to do so they had
to step on the mat and in consequence, their body weight prevented the cart from mov-
ing. In the first experiment, performance on the shopping cart task was examined in
children of 16 and 21 months both when the self was the obstacle and when a heavy
weight was the obstacle. Results showed significant improvement with age in perfor-
mance for the self-version of the task that was not matched by similar age differences in
the weight task. In the second experiment, children’s performance on both the self-
version of the shopping cart and on a standard mirror self-recognition task was assessed.
Results showed a significant correlation independent of age for these 2 tasks. These find-
ings provide further evidence for the notion that toddlers develop an objective awareness
of the self during the 2nd year. They are also consistent with the idea that body self-
awareness is heightened in situations requiring self-monitoring during movement.
For more than 100 years, psychologists have made a distinction between aware-
ness of the first-person aspects of the self, what James (1890/1950) called the I,
and awareness of the self as an object, the Me. It is generally accepted that some
Correspondence should be addressed to Chris Moore, Department of Psychology, Dalhousie
University, Halifax, Nova Scotia, B3H 4J1, Canada. E-mail: moorec@dal.ca
Downloaded by [Dalhousie University] at 05:56 22 February 2012
elements of first-person experience corresponding to the self are in place from
early in the first year of life. For example, Bahrick and Watson (1985) demon-
strated that infants from 3 to 5 months were able to detect the correspondence
between visual and proprioceptive information of their own movements in the
sense that they could extract an amodal temporal invariance when presented with
live video of their own leg movements. There is also evidence that such young
infants can also extract spatial invariances across visual and proprioceptive infor-
mation (Rochat & Morgan, 1995).
In contrast to this early developing sensitivity to first-person information, the
awareness of the self as an individual objective entity is generally believed to
occur around the middle of the second year. For the last 30 years, the main
approach to the study of objective self-awareness has been mirror self-recognition.
This approach was first developed as a nonverbal method to study self-recognition in
chimpanzees by Gallup (1970). Gallup anesthetized the chimpanzees and applied
red dye to different parts of their bodies. When the animals recovered from the
anesthetic, they were placed in front of a mirror. Gallup reported that under these
conditions, chimpanzees would explore the red marks and in particular would use
the mirror to explore marks on their faces. A similar procedure was used indepen-
dently with young children by Amsterdam (1972), who used rouge to mark the
cheeks of children from 3 to 24 months and then recorded their behavior in front
of a mirror. Amsterdam reported that late in the first year and early in the second,
children’s most common response was to show a combination of social behavior
to the mirror image, searching for the image in or behind the mirror, and observ-
ing the effects of their own movement in the mirror. Only after 18 months did
children start to show evidence of recognizing the image as the self by touching
the mark on their own faces.
This basic pattern using the marked face procedure has now been replicated
many times (e.g., Bertenthal & Fischer, 1978; Johnson, 1982; Lewis & Brooks-
Gunn, 1979; Nielsen, Dissanayake, & Kashima, 2003). It has also been extended
to the recognition of a marked leg (Neilsen, Suddendorf, & Slaughter, 2006).
Whereas the developmental pattern of behavior is quite well agreed on, there is
still debate over what the behavior means. Some have argued that mirror self-
recognition reveals little more than the learning of the reflective properties of
mirrors and the use of mirrors to explore the body (e.g., Loveland, 1986).
However, most authors agree that mirror self-recognition does tell us something
interesting about the development of the self. In particular, it is generally
believed that mirror self-recognition evidences children’s representations of their
appearance, whether more permanent as in facial appearance or relatively tempo-
rary, as in clothing (e.g., Neilson et al., 2006). In this way, facial appearance
observed in the mirror is recognized by the child to be consistent with the child’s
mental image of himself or herself. Therefore, when the child sees the marked face
in the mirror, the child immediately knows that the mark is on his or her own face.
158 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
Mirror self-recognition shows, then, that the child can imagine self from an
objective point of view—as me.
It has also been claimed that mirror self-recognition depends on the child’s
ability to process the identity relation between the visual information provided by
the mirror and the proprioceptive information provided by the child’s movements
while in front of the mirror (Mitchell, 1993; Povinelli, 1995). It is important to
note that processing the identity relation is more than just extracting the amodal
temporal invariance for particular movements across proprioceptive and visual
modalities, which, as we have mentioned, even very young infants can do
(Bahrick & Watson, 1985). It means understanding that the two forms of informa-
tion are equivalent (Mitchell, 1997; Povinelli, 1995). Given this equivalence,
children can determine that what is true of the image in the mirror is also true of
the self and therefore that the mark exists on their own face.
The mirror self-recognition task has been a productive research tool for a
number of years; however, in that it is tied to the recognition of visual appearance, it
is necessarily an assessment of only a limited aspect of the objective self. Clearly
there are other aspects of the self and a greater variety of methods would be desir-
able if one wishes to assess the extent to which children acquire a general under-
standing of the objective nature of the self. One other construal of the self that is of
interest to us here is the awareness of the self as a body with physical properties.
There is preliminary evidence that body awareness emerges at the end of
infancy. Piaget (1953/1977) recorded the following observation of his daughter
Jacqueline at 18.5 months: “Jacqueline is standing on a rag (50 × 30 cm.), which
she is trying to pick up. She pulls, is surprised at the resistance, but it does not
occur to her to move. Finally she gives up” (p. 351, obs. 168). Within a month she
was able to solve such a problem. Piaget interpreted this observation and others
(e.g., Piaget, 1954) in which his children attempted to negotiate the spatial envi-
ronment and the causal relations among objects in that environment in terms of a
developing awareness of self as an objective entity existing in, and interacting
with, the world of objects. For Piaget, the objectification of the bodily self occurs
as the infant becomes able to represent the body’s spatial and causal relations
with the external world (Piaget, 1954). Two studies have followed up Piaget’s
observations. Geppert and Küster (1983) and Bullock and Lütkenhaus (1990) sat
toddlers on a small blanket and then asked the children either simply to give the
blanket to the experimenter or to retrieve an attractive object hidden underneath
the blanket. Most children were able to pass this task by 18 months. Although it is
intriguing that children seem to be able to perform well on the blanket task at
about the same time as they pass the mirror self-recognition task, it is unclear to
what extent their success was the result of learning from similar situations that
may have occurred regularly in natural circumstances. It is quite possible that
infants have regularly found themselves sitting or standing on a blanket or cloth
that they wish to retrieve.
BODY SELF-AWARENESS 159
Downloaded by [Dalhousie University] at 05:56 22 February 2012
In this study, we attempted to design a task involving novel circumstances
that would test children’s awareness of themselves as an objective entity with
physical properties. In designing this task, we also wanted to provide a test under
conditions in which the child was locomoting. There is a theoretical reason for
believing that body awareness may be best elicited under conditions of locomo-
tion. To understand why, it is necessary to consider how self-awareness may
have evolved. There is strong evidence that mirror self-recognition is limited to
the hominoid family. Whereas members of all great ape species have shown evi-
dence of mirror self-recognition, monkeys do not (Povinelli, 1995). In the only
existing explanation of why self-awareness evolved in the hominoid line,
Povinelli and Cant (1995) argued that the particular ecological challenge of
being a large-bodied primate in an arboreal environment resulted in an adapta-
tion whereby the arboreal common ancestor of modern-day apes was able to
construct an integrated representation of its body in relation to the environment.
Based on their observations of arboreal locomotion of orangutans, Povinelli and
Cant suggested that the particular challenge arose from the difficulty of moving
a large and heavy body through an environment that was relatively unstable. The
arboreal physical environment tends to be much more unpredictable in terms of
support than the terrestrial physical environment. Under the weight of a large
body, branches move and deform in unpredictable ways. They may break, lead-
ing to sudden loss of support. As a result, large-bodied arboreal apes cannot rely
on stereotyped locomotion programs; rather, they need to constantly monitor the
relation between all of their limbs and the environment to avoid the potentially
severe negative fitness consequences of falling. In contrast, relatively small-bodied
arboreal animals, including monkeys, are able to traverse their environment with
relative impunity and so are able to rely on relatively stereotyped movement pat-
terns in much the same way that humans do on flat surfaces once they have
learned to walk or run. This is not the place to consider whether the overall
paucity of relevant evidence from primate species warrants advancing an evolu-
tionary account such as Povinelli and Cant’s (1995) clambering hypothesis for
the evolution of self-awareness (de Veer & van den Bos, 1999). Suffice to say
that if they are correct, then one prediction is that self-awareness might be more
likely to be elicited in situations involving a challenge to locomotion (see also
Barth, Povinelli, & Cant, 2004).
Given this background, we designed a task that we believed would present
such a challenge to locomotion. The task requires the child to push a toy shop-
ping cart. Ordinarily, such a task would be straightforward for toddlers who are
already walking without support. The difference is that our shopping cart is
altered so that a small mat is attached to the back axle. The mat can be laid out to
extend about 15 in. behind the handle of the cart (see Figure 1). If the child is to
160 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
push the cart in the canonical manner, he or she must step onto the mat to grasp
the handle. As a result, the child’s weight impedes the forward motion of the cart.
The child’s pushing creates a force on the underside of the child’s feet as the mat
is pulled. The mat can also be folded so that it sits entirely under the handle, and
in this arrangement, the child does not have to step onto the mat to push. At the
same time, an independent weight can be placed on the folded mat to cause a
similar impediment to forward motion.
In the first study reported here, we examined in a cross-sectional experiment
whether there would be a difference in performance on the shopping cart task for
children on either side of the 18-month transition. We predicted that older toddlers
would show superior performance on the shopping cart task when compared to
younger toddlers. Alone such a result would not be surprising—children improve
on most measures of performance with age. Because our shopping cart task might
be construed to require children to reason about the causal relations involved, a gen-
eral development in causal (or mean–ends) reasoning might be responsible for
performance improvements independently of any change in bodily self-awareness.
To examine this issue we compared performance on self and object versions of the
task. In the self-version, the child’s own body acted as the impediment, whereas in
the object version, a paint can filled with sand acted as the impediment.
In the second study, to further examine the idea that the shopping cart might
tap into the development of awareness of self as an objective entity, we explored
whether performance on the shopping cart task was related to performance on
mirror self-recognition.
BODY SELF-AWARENESS 161
FIGURE 1 Toy shopping cart with (a) mat extended for self condition, and (b) paint can in
place for object condition.
Downloaded by [Dalhousie University] at 05:56 22 February 2012
STUDY 1
Method
Participants
Eighteen 15-month-olds (8 girls and 10 boys) and twenty-one 21-month-olds
(12 girls and 9 boys) from a small Canadian city participated in this study. All
participants were within 1 month either side of their 15-month or 21-month birth-
day. Participants were recruited through birth announcements in local newspa-
pers. All children were able to walk unassisted.
Material and Apparatus
The main apparatus used for this study was a child’s plastic shopping cart (52
cm high to the handle). A pink-and-white checked mat was attached to the back of
the shopping cart by means of string looped around the back legs of the shopping
cart. The mat trailed behind the cart. Strips of Velcro were attached to the mat for
ease of rolling up during the object trials. A 1-gallon paint can that had been
painted white was used during the object trials. The paint can was full of sand and
weighed about 5 kg. It was mounted on four furniture castors to allow greater ease
of movement.
Procedure
On arrival at the lab, the participant was shown a variety of toys and was allowed
to play with them to become familiarized with the lab setting and with the experi-
menter. After the experimenter and the parent judged that the participant was suffi-
ciently accustomed to the lab setting and to the experimenter, the toys were put
away and the experiment began. The whole session was videotaped for later coding.
Children in each age group were given both experimental tasks in a counterbal-
anced order. For purposes of simplicity, the self-task followed by the object task is
presented here. After the toys had been put away, the experimenter approached the
cart, which had been placed out of the participant’s view, and placed the cart on a
mark on the floor, with the mat laid out flat behind it. The experimenter then pro-
ceeded to demonstrate to the participant how to roll up the mat as well as how to
push the cart. Following this, the experimenter allowed the participants to push the
cart themselves with the mat rolled up so they would not step on it.
After familiarization with the cart and the mat, the cart was returned to the
starting position, with the mat again trailing behind it. The parent was seated
approximately 6 ft away. The experimenter placed the participant behind the cart
and the mat. The experimenter then told the participant to “push the cart to
mommy/daddy.” The parent was also allowed to ask for the cart, but was instructed
162 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
not to give his or her child any information on how to complete the task. To reach
the handle of the shopping cart, participants had to step onto the mat, thus
preventing them from pushing the cart. Each trial continued until the participant
successfully pushed the cart to his or her parent or the participant had walked
away from the cart and did not display any further interest in the cart. The same
procedure was repeated for a maximum of four separate trials, each one begin-
ning with the participant being placed behind the cart and the mat.
For the object task, the mat was rolled up by the experimenter and fixed in
place with Velcro, and the paint can was placed on the mat in place of the child’s
weight. At the beginning of each trial, the participant was placed directly behind
the cart, and again asked to “push the cart to mommy/daddy.” The parent was
allowed to ask for the cart as well, as in the self-trials, and each trial continued
until the participant successfully pushed the cart, or walked away and displayed
no further interest in the cart. The same procedure was repeated for a maximum
of four object trials.
Not all children completed all four trials of each task. For the self-task, 31
children completed all four possible trials, 4 completed three trials, and 2 children
each completed only two and one trials. For the object task, 27 children
completed all four trials, 9 completed three trials, and 3 completed two trials.
Coding and Measures
Primary coding was carried out by one coder who was otherwise uninvolved in
the study and was naive to its purposes. She coded all participants on each com-
pleted trial for the following categories.
Successes.
A success was scored if the child succeeded in moving the
shopping cart at least one cart length toward the parent. This criterion was chosen
because it meant the child had succeeded in moving the cart even if it was not
pushed all the way to the mother. Success could not be achieved without stepping
off the mat in the self-task or moving the paint can off the mat in the object task.
In the self-task, success could be achieved by stepping to the side of the cart and
pushing, moving to the front of the cart and pulling, or moving the mat out of the
way and pushing. The most common strategy was to push or pull from the side or
front. Only rarely was the mat moved out of the way. In the object task, success
was typically achieved by pushing the paint can, which usually led to it tipping
over and off the mat. Once the child had tried to move the paint can the experi-
menter would assist in moving it off the mat so that the child did not get hurt.
When the paint can was off the mat, the cart could be pushed forward in the
canonical manner. Each completed trial was scored dichotomously as either a
success or not.
BODY SELF-AWARENESS 163
Downloaded by [Dalhousie University] at 05:56 22 February 2012
Attempts.
An attempt was coded if the child placed both hands on the han-
dle of the shopping cart and appeared to push. The form of attempts was the same
for both self- and object tasks. All such attempts were counted for each trial so
there could be more than one attempt per trial.
Reliability coding was carried out on 14 participants by a second coder also
naive to the purposes of the experiment. For both successes and attempts reliabil-
ity was good (r= .88 and .93, respectively).
Because not all children completed four trials for both self- and object tasks,
measures of successes and attempts in both tasks were generated by dividing the
successes and attempts raw scores by number of trials completed. For conve-
nience, these two measures are referred to in what follows as successes and
attempts. The success measure was a proportional measure with a possible range
from 0 to 1. The attempts measure was the rate of attempts per trial completed. In
addition, the raw attempts and successes measures were combined into a measure
of overall performance efficiency across all trials of each task by dividing the
total number of successes by the number of successes plus the number of
attempts. The denominator here represents the total number of efforts the children
had at pushing the cart, whether successful or not. Thus, this measure reflects in
part the degree to which children engaged in trial and error. A high efficiency
score reflects success with relatively few attempts and so may be more reflective
of insightful problem solving. Thus for each task (self and object) there were
three measures—successes, attempts, and efficiency.
Results
First, performance on each measure was compared to examine sex effects.
Repeated measures analyses of variance (ANOVAs) with task as the within-
subjects variable and age group and sex as between subject variables were run.
The only significant sex effect was that boys (M= .55) had overall more successes
than girls (M= .36), F(1, 35) = 5.76, p < .05, η2= .141. Attempts and efficiency
did not show sex effects nor did sex interact significantly with age. Given that
successes at least on the object task could be due in part to brute strength and that
neither attempts nor efficiency differentiated boys and girls, sex was not consid-
ered further in the analyses.
Preliminary analysis was also conducted examining the number of trials of
each task completed as a function of task order. This analysis showed a significant
interaction of task and order, F(1, 37) = 4.42, p< .05, η2= .107). Children com-
pleted more self-trials than object trials when the self-task was first and more
object trials than self-trials when the object task was first. Consequently, the two
task orders were examined separately in all subsequent analyses.
For each order, the three main types of variables were analyzed with task (self,
object) by age group repeated measures ANOVAs. The same analyses were also
164 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
carried out on raw frequencies of successes and attempts and the results did not dif-
fer. Means and standard deviations are shown in Table 1. When children received
the self-task first, the analysis of successes showed a main effect of task, F(1, 18) =
6.53, p< .05, η2= .266, and a main effect of age, F(1, 18) = 9.95, p< .01, η2= .356.
Children were more successful on the self-task than the object task in general and
the older children were more successful than the younger children. The interaction
was not significant, F(1, 18) = 2.13, ns,η2= .106. In contrast, when children
received the object task first, there were no significant differences in success.
The analyses of attempts showed that when children received the self-task
first, there were no main effects but a significant interaction between task and age,
F(1, 18) = 4.77, p< .05, η2= .209. Younger children made more attempts than
older children on the self-task but fewer attempts than older children on the object
task. When the object task was first, there were no significant effects at all.
The analyses of efficiency revealed a significant interaction between task and
age when the self-task came first, F(1, 18) = 5.56, p< .05, η2= .236. Older
children were significantly more efficient than younger children on the self task
(t= 3.29, p< .01, d= 1.47), but there was no significant difference between
younger and older children on the object task. When the object task came first,
there were no significant differences in efficiency.
Finally, correlations between the two tasks were run for each measure for each
age group. None of these correlations were significant. In particular, for the self
and object efficiency scores, the correlation was r= .032 for the younger group
and r= .140 for the older group, indicating the relative independence of perfor-
mance on the two types of task.
BODY SELF-AWARENESS 165
TABLE 1
Means and Standard Deviations for Success, Attempts, and Efficiency on Each Task
as a Function of Age Group and Task Order in Study 1
Success Attempts Efficiency
Self Object Self Object Self Object
Task M SD M SD M SD M SD M SD M SD
Self task first
Younger .342 .327 .308 .275 3.17 1.66 1.91 0.84 .099 .113 .113 .202
Older .850 .269 .575 .409 2.30 1.23 3.23 1.77 .308 .166 .122 .181
Object task first
Younger .219 .364 .291 .388 1.89 0.85 2.33 0.99 .160 .203 .123 .170
Older .349 .450 .636 .438 2.04 1.45 2.50 1.64 .152 .125 .256 .202
Note. Success was a proportion of number of trials completed that were solved and had a possi-
ble range from 0 to 1. Attempts was a rate measure. Efficiency was calculated according to the fre-
quency of successes/successes + attempts.
Downloaded by [Dalhousie University] at 05:56 22 February 2012
Discussion
This study was contrived to examine age differences in a task designed to tap bod-
ily awareness. The self-version of the shopping cart task, conceived from ideas
stemming both from Piaget’s (1953/1977) observations and Povinelli and Cant’s
(1995) hypothesis of the origins of bodily self-awareness required children to
locomote in a situation in which their own body would impede their success. The
results showed that the older group of children, who were 21 months of age, were
much more successful than the younger group, who were 16 months of age. This
effect was seen most strongly for the condition where the self-version of the task
was presented first. Performance was generally poorer when the self-task fol-
lowed the object task, presumably because of fatigue and loss of interest. Indeed
children completed fewer self-trials when this task was presented second and 2 of
the children only completed one self-trial in this task order. Because of this order
effect, it is important to pay most attention to performance when the self-task was
presented first. Under these circumstances, the older children were successful at
moving the cart forward on an average of 3.3 trials out of a maximum of 4, com-
pared to the younger children who were successful on only 1.3 trials. At the same
time the older children achieved this success with rather fewer attempts at push-
ing the cart. The combined efficiency measure shows that the success rate per try
was much higher for the older group than for the younger group. This efficiency
measure is the clearest evidence that the older children were much better able
than the younger children to understand the demands of the task and solve it. It is
perhaps worth mentioning, however, that not even the older children achieved
efficiency anywhere close to 1. Thus, all children appeared to solve the task
largely by trial and error. In short, there was no evidence of truly insightful prob-
lem solving in this task.
So far, this result is perhaps of little significance. After all, older children gen-
erally do better than younger children at most tasks. The older children had con-
siderably more experience locomoting in general and so should be expected to
perform better on tasks requiring locomotion. To determine whether the superior
performance displayed by older children might be related to developing self-
awareness, we also used a control task—the object task—formally equivalent to
the self-task but not requiring any reflective self-awareness for success.
Examination of performance on the object task when it was presented first
showed no significant differences across age in success at moving the cart for-
ward, attempts at moving the cart, or overall efficiency. These results suggest that
the age-related performance seen on the self-task was not due to more general
aspects of development, such as a change in general locomotory skill or in the
ability to reason about the causal relations involved in the task. Rather it is some-
thing more specific to do with children’s changing awareness of their own bodies
in such circumstances. Although it is a null result, the finding that performance in
166 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
the self-task was not correlated with performance in the object task further
supports the inference that the two tasks to some extent tap different abilities.
This relative independence of performance on the self- and object tasks might
seem counterintuitive at one level. Certainly, according to a traditional Piagetian
(1954) account, it might have been predicted that performance on the two tasks
would be more closely related because both require cognitive abilities emerging
at the end of the sensorimotor period. In particular, as children become able to
consider the independent roles that objects can take up in spatial and causal rela-
tions, they should also become able to consider the self as one possible object in
such relations. However, our results suggest that self-awareness is separate
enough developmentally from more general aspects of spatial and causal reason-
ing to yield differentiable effects despite quite similar task demands.
Having shown that there were age-related differences in performance on the self-
version of the shopping cart task, a second study was designed to explore further the
idea that successful performance on this task taps into self-awareness. In this study,
we examined children within the age range for their performance on both shopping
cart performance and on a more standard measure of self-awareness—the mirror
self-recognition task. The rationale was that if both tasks are valid measures of a
developing awareness of the objective self then performance on the two tasks dur-
ing the critical transitional age range should be correlated (see also Müller &
Liebermann, 2004).
STUDY 2
Method
Participants
Twenty-nine children from the same population as in Study 1 participated in
this study. Children ranged in age from 16 to 24 months (M= 18.27, SD = 2.36).
Participants were primarily from White, middle-class families. One child from
the original 29 was dropped from the experiment because she could not yet walk;
one other child did not complete both tasks. For the final sample of 27, children
ranged in age from 16 to 24 months (M= 18.32, SD = 2.39).
Procedures and Measures
On arrival in the laboratory, the participants were allowed to become familiar
and comfortable with the setting. During this time the experimenter played with
the child with some toys and the sessions began after the child was at ease.
Children were given the shopping cart task first and the mirror self-recognition
task second. Both tasks were videotaped for later coding.
BODY SELF-AWARENESS 167
Downloaded by [Dalhousie University] at 05:56 22 February 2012
Shopping cart task.
The procedure for the shopping cart task was the same
as in Study 1. For this experiment, the task ended when the child completed two
successful pushes or completed four trials, whichever came first. The choice to
end after two successes was made because there was some indication from our
earlier work that children who were able to succeed immediately on the shopping
cart task were less inclined to continue for four trials. A successful push was
coded when a child pushed the shopping cart one full length of the cart. An
attempt was coded when the child stood on the carpet and tried to push.
Mirror recognition task.
The mirror self-recognition tasks followed a stan-
dard approach. After the shopping cart task was completed, a round sticker was
placed on the child’s forehead using the pretext of wiping some dirt from his or
her forehead. The child was then placed in front of a mirror (65 cm high, 48 cm
wide). No children attempted to remove the sticker before being placed in front of
the mirror. After approximately 30 sec, if the child had not yet removed the
sticker, the experimenter attempted to elicit self-exploration using the following
series of prompts, leaving between each prompt an opportunity for the child to
respond: “Who’s that? What’s that (the experimenter pointed to the sticker in the
mirror)? Where is that sticker really?”
Measures.
Coding of the shopping cart task was carried out in essentially
the same way as for Study 1. Success scores were computed as proportions of tri-
als completed. Attempts were computed as rates per trial. All coding was done by
an individual who was blind to the purpose of the study. A second coder rated a
random sample of 10 participants to assess interrater reliability. For both suc-
cesses and attempts reliability was good (r = .86 and .95, respectively). For the
mirror self-recognition task, the same coders examined whether and at what point
the children reached up to their forehead. Reaching for the sticker before the
prompts was coded as 2, children who reached up in response to the prompts were
coded as 1, and children who did not reach up were coded as 0. This three-level
coding was carried out to tap additional variability in performance on the grounds
that some children in the age range studied might be transitional in mirror self-
recognition and might only show it when explicitly prompted (or scaffolded) to
do so. Interrater agreement for mirror self-recognition was perfect.
Finally, as for Study 1, an efficiency score reflecting both attempts and success
was calculated (efficiency = successes /attempts + successes).
Results and Discussion
The mean scores for each measure were as follows: attempts, M= 1.33, SD =
1.27; successes, M= .47, SD = .38; efficiency, M= .30, SD = .28; mirror
168 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
self-recognition, M= 1.14, SD = .85. For the mirror self-recognition task, 8
children completely failed to touch the sticker, 8 touched the sticker only after the
prompt, and 11 touched the sticker when first shown the mirror before the
prompt. Performance on both self-tasks was consistent with previous research on
children of around 18 months. There was considerable variability in performance
on the shopping cart task and this was consistent with Study 1, which found that
there was considerable improvement on this task between 16 and 21 months of
age. This study also found variability in mirror self-recognition consistent with
the well-replicated finding that children typically start to show self-directed
behavior when they observe their own marked face in a mirror at about 18 months
of age (e.g., Bertenthal & Fischer, 1978; Johnson, 1982; Lewis & Brooks-Gunn,
1979; Nielsen et al., 2003). So this study on children averaging 18 months yielded
typical performance on two tasks assessing aspects of the objective self.
Bivariate correlations were performed for all variables including age. Given the
clear-cut prediction made at the outset, a one-tailed test of significance was used
and revealed that age correlated significantly with successes and mirror self-
recognition but not with attempts and efficiency. Mirror self-recognition correlated
significantly with efficiency (r= .411, p< .05) and successes (r= .375, p < .05).
Table 2 presents the bivariate correlations.
Because significant correlations between two variables may be obtained through
their common associations with more general underlying developmental factors,
such as cognitive development, we next carried out correlational analyses between
performance on the two tasks of self-awareness with age controlled. For all three
shopping cart task measures (attempts, successes, and efficiency), partial correlations
with the mirror self-recognition score were calculated, with age in months partialed
out. Using a one-tailed test of significance, only efficiency was significantly corre-
lated with mirror self-recognition (r= .35, p< .05). Inspection of the descriptive
means in Table 3 indicates that children who scored 0 on the mirror self-recognition
BODY SELF-AWARENESS 169
TABLE 2
Correlation Matrix of Measures for Study 2
Variable 2a3b4b5b
1. Age in months .36* .22 .43* .26
2. Mirror score — –.10 .38* .41*
3. Successes — .09 .79**
4. Attempts — –.43*
5. Efficiency —
aN=28. bN=27.
*p<.05, **p< .01, one tailed.
Downloaded by [Dalhousie University] at 05:56 22 February 2012
task had a lower efficiency score than children who scored 1 and 2. As well,
children who scored 2 had a higher score than those who scored 0 or 1.
Table 3 also indicates an interesting pattern for the other two measures on the
shopping cart task. Children who scored 0 on the mirror self-recognition task
tended to have few attempts as well as few successes (Mattempts = 1.28, Msuc-
cesses = .25). Children who scored 1 on the mirror self-recognition task made the
most attempts and had more successes than those who scored 0 (Mattempts =
1.78, Msuccess = .51). Children who scored 2 on the mirror self-recognition task
had the most successes, but also made the fewest attempts (Mattempts = 1.05, M
successes = .59). This pattern suggests that those children who were perhaps tran-
sitional on the mirror self-recognition task were able to solve the shopping cart
task, but did so in a more trial-and-error fashion than the children who passed the
self-recognition task without prompts.
Finally, as a more conservative test of the association between mirror self-
recognition and shopping cart performance, we carried out univariate ANOVAs
on the three measures of shopping cart performance—successes, attempts, and
efficiency—with mirror self-recognition coded dichotomously as pass (touched
the sticker either before or after the prompt) or fail (did not touch the sticker).
Mirror self-recognition passers had more successes than failers, F(1, 25) = 4.25,
p= .05, η2=.15, and passers were more efficient than failers, F(1, 25) = 4.28, p=
.049, η2= .15. There was no difference between passers and failers in attempts to
push the shopping cart.
The novel contribution of this study is simple. Performance on the two tasks
used in this study was significantly correlated independently of age. In particular,
performance on the mirror self-recognition task was significantly correlated with
the shopping cart task efficiency measure, which represents how easily partici-
pants were able to solve the shopping cart task. It should be noted that, perhaps as
a result of a relatively small sample size, the magnitude of the correlation was not
170 MOORE, MEALIEA, GARON, POVINELLI
TABLE 3
Means for the Three Shopping Cart Measures as a Function of Mirror
Self-Recognition Score in Study 2
Mirror Self- Shopping Cart Shopping Cart Shopping Cart
Recognition Score Success Attempt Efficiency
0 .250 1.28 .154
1 .511 1.78 .376
2 .592 1.05 .496
Note. Success was a proportion of number of trials completed that were solved and had a possi-
ble range from 0 to 1. Attempts was a rate measure. Efficiency was calculated according to the fre-
quency of successes/successes +attempts.
Downloaded by [Dalhousie University] at 05:56 22 February 2012
huge and was only significant using a one-tailed test. Nevertheless, this result
implies that there is a developmental link between the two measures even though
at face value they differ considerably. The link we suggest is that both evidence
children’s developing awareness of the objective status of the self. In the final
section, we consider in more detail what this awareness entails.
GENERAL DISCUSSION
We have developed a new task of bodily self-awareness that reveals developmen-
tal change at about 18 months. This task requires children to be aware of them-
selves as objects that can interact physically with other objects in the world. The
task presents children with a novel situation and so has the advantage that it is
very unlikely that children will have acquired a solution method prior to testing.
Younger children find the task difficult and when they do solve it, they appear to
do so by trial and error. Older children solve the task more often and solve it with
fewer attempts.
We believe that this task is a valid assessment of the understanding of self as an
object. First, it is the objective physical nature of the actor’s body that impedes
forward motion on the shopping cart. Therefore, efficient solution of the task
implies that the actor recognizes this characteristic. Second, we have convergent
validity in the form of a significant age-independent correlation between perfor-
mance on the shopping cart task and performance on the standard mirror self-
recognition task. It is important to note that this correlation was moderate,
implying that the two tasks tap related but different aspects of the objective self.
Recently Brownell and colleagues (Brownell, Zerwas, & Ramani, 2006) also
used a range of similar tasks with toddlers between 18 and 30 months and pro-
vided evidence for relatively dissociable aspects of the objective self. However, it
is also noteworthy that the correlation between the two tasks in Study 2 was rather
larger than that between the two versions of the shopping cart in Study 1, even
though the latter two tasks present much more similar task demands and might be
imagined to depend on a common understanding of causal relations.
Many authors have recognized the multifaceted nature of the self (e.g., James,
1890/1950; Lewis & Brooks-Gunn, 1979). Sensitivity to the first-person informa-
tion specifying the self is clearly present early in infancy (e.g., Bahrick & Watson,
1985). However, the representation of the self from an objective or third-person
point of view—the Me—develops later during the second year of life. But what
counts as having an objective sense of self? On the one hand, representing self
from an objective point of view entails understanding that the self is a unique
object. The mirror self-recognition task reveals young children’s awareness of the
self as a particular entity—the self is a person who can be differentiated from
other people based on visual appearance. On the other hand, representing the self
BODY SELF-AWARENESS 171
Downloaded by [Dalhousie University] at 05:56 22 February 2012
objectively entails recognizing that the self is also an object like other objects—it
has physical properties that play a causal role in the world of objects. The shop-
ping cart task assesses this aspect of the self.
If we are correct that the objective self comes to be understood both as a
unique entity and also as an object like other objects, can we account for how
these two facets of the objective self develop in close synchrony around the mid-
dle of the second year? Both of these facets require the child to imagine informa-
tion about the self that is not typically available through direct perception
(third-person information about the self) and then to relate it to information that is
directly available through perception (first-person information, in particular via
proprioception). Like other authors (e.g., Perner, 1991; Suddendorf & Whiten,
2001), we believe that more general developmental changes in representational
ability play a role here. Around the middle of the second year, children become
capable of holding in mind a secondary representation as they are engaged per-
ceptually with the environment (primary representation). This capability is evi-
denced by a variety of cognitive and social achievements, including aspects of
object representation, pretend play, and imitation (e.g., Lewis & Ramsey, 2004;
Nielsen & Dissanayake, 2004; Suddendorf & Whiten, 2001). The capacity for
secondary representation enables relations between perceived and imagined
information to be established. The self as an entity with an objective nature is one
conceptual outcome of this level of information processing.
Although recognizing the developmental role of the onset of secondary repre-
sentation, it is also important to point out that the concept of the objective self
requires the integration of the two sources of information about the self that
appears to be facilitated under conditions of movement. Both Mitchell (1993) and
Povinelli (1995) argued that mirror self-recognition depends on an awareness of
the correspondence between the visual information of the self’s movement in the
mirror and the proprioceptive information provided from the same movement.
Indeed if that correspondence is removed while maintaining visual information of
the self’s appearance by using delayed video, young children fail to show self-
directed behavior (Povinelli, Landau, & Perilloux, 1996). Thus it is the monitor-
ing of the contingent relation between the self’s action and the visually apparent
motion in the mirror that is essential for mirror self-recognition. The shopping
cart task also presents children with a situation in which they must monitor the
relation between their own movement and the movement (or lack of it) of the
shopping cart that they are controlling. To solve the task, the children must recog-
nize that their own body is the cause of the lack of effective movement. Arguably,
then, this task also requires children to process and represent the contingent
motion relations between self and world. The fact that even the older children in
Study 1 solved the self-version of the shopping cart largely through trial and
error rather than through insight is consistent with this idea. It is possible that this
situation harkens back to the time when the earliest hominoid ancestors first were
172 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012
confronted with the need to represent the movement of the objective self in
relation to the environment (Povinelli & Cant, 1995).
ACKNOWLEDGMENTS
This research was conducted at Dalhousie University with support from a grant
from the Social Sciences and Humanities Research Council of Canada. Our
thanks go to Carly McCreath, Kaja Montgomery, and Natasha Wirtanen for their
assistance with the project. Thanks also to Celia Brownell, Ulrich Mueller, and an
anonymous reviewer for feedback on an earlier draft of this article.
REFERENCES
Amsterdam, B. (1972). Mirror self-image reactions before age two. Developmental Psychobiology,
5, 297–305.
Bahrick, L. E., & Watson, J. S. (1985). Detection of intermodal proprioceptive visual contingency as a
potential basis of self-perception in infancy. Developmental Psychology, 21, 963–973.
Barth, J., Povinelli, D., & Cant, J. (2004). Bodily origins of SELF. In D. Beike, J. Lampinen, &
D. Behrend (Eds.), The self and memory (pp. 11–43). New York: Psychology Press.
Bertenthal, B., & Fischer, K. (1978). Development of self-recognition in the infant. Developmental
Psychology, 14, 44–50.
Brownell, C. A., Zerwas, S., & Ramani, G. B. (2006). “So big”: The development of bodily self-
awareness in toddlers. Manuscript submitted for publication.
Bullock, M., & Lütkenhaus, P. (1990). Who am I? Self-understanding in toddlers. Merrill-Palmer
Quarterly, 36, 217–238.
de Veer, M., & van den Bos, R. (1999). A critical review on methodology and interpretation of mirror
self-recognition in non-human primates. Animal Behaviour, 58, 459–468.
Gallup, G. G. (1970). Chimpanzees: Self-recognition. Science, 167, 86–87.
Geppert, U., & Küster, U. (1983). The emergence of “wanting to do it oneself”: A precursor of
achievement motivation. International Journal of Behavioral Development, 6, 355–369.
James, W. (1950). The principles of psychology (Vol. 1). New York: Dover. (Original work published 1890)
Johnson, D. (1982). Altruistic behavior and the development of the self in infants. Merrill-Palmer
Quarterly, 28, 379–388.
Lewis, M., & Brooks-Gunn, J. (1979). Social cognition and the acquisition of self. New York: Plenum.
Lewis, M., & Ramsay, D. (2004). Development of self-recognition, personal pronoun use, and pretend
play during the 2nd year. Child Development, 75, 1821–1831.
Loveland, K. A. (1986). Discovering the properties of a reflecting surface. Developmental Review,
6, 1–24.
Mitchell, R. W. (1993). Mental models of mirror self-recognition: Two theories. New Ideas in
Psychology, 11, 295–325.
Mitchell, R. (1997). Kinesthetic-visual matching and the self-concept as explanations of mirror self-
recognition. Journal for the Theory of Social Behavior, 27, 17–39.
Müller, U., & Liebermann, D. (2004). Facing the body: Toward a developmental theory of body
knowledge. Monographs of the Society for Research in Child Development, 69(Serial No. 276),
103–113.
BODY SELF-AWARENESS 173
Downloaded by [Dalhousie University] at 05:56 22 February 2012
Nielsen, M., & Dissanayake, C. (2004). Pretend play, mirror self-recognition and imitation:A longitu-
dinal investigation through the second year. Infant Behavior and Development, 27, 342–365.
Nielsen, M., Dissanayake, C., & Kashima, Y. (2003). A longitudinal investigaton of self–other dis-
crimination and the emergence of minor self-recognition. Infant Behavior and Development, 26,
213–226.
Nielsen, M., Suddendorf, T., & Slaughter, V. (2006). Mirror self-recognition beyond the face. Child
Development, 77, 176–185.
Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT Press.
Piaget, J. (1954). The construction of reality in the child. New York: Basic Books.
Piaget, J. (1977). The origin of intelligence in the child. Harmondsworth, England: Penguin. (Original
work published 1953)
Povinelli, D. P. (1995). The unduplicated self. In P. Rochat (Ed.), The self in early infancy (pp. 161–192).
Amsterdam: North Holland-Elsevier.
Povinelli, D. P., & Cant, J. (1995). Arboreal clambering and the evolution of self-conception. The
Quarterly Review of Biology, 70, 393–421.
Povinelli, D. P., Landau, K., & Perilloux, H. (1996). Self-recognition in young children using delayed
versus live feedback: Evidence of a developmental asynchrony. Child Development, 67, 1540–1554.
Rochat, P., & Morgan, R. (1995). Spatial determinants in the perception of self-produced leg move-
ments in 3- to 5-month-old infants. Developmental Psychology, 31, 626–636.
Suddendorf, T., & Whiten, A. (2001). Mental evolution and development: Evidence for secondary
representation in children, great apes, and other animals. Psychological Bulletin, 127, 629–650.
174 MOORE, MEALIEA, GARON, POVINELLI
Downloaded by [Dalhousie University] at 05:56 22 February 2012