Siblings, executive functioning and theory of mind

Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
733
British Journal of Developmental Psychology (2006), 24, 733–751
q 2006 The British Psychological Society
The
British
Psychological
Society
www.bpsjournals.co.uk
Mental playmates: Siblings, executive functioning
and theory of mind
Anna McAlister and Candida C. Peterson*
University of Queensland, Australia
This study assessed the theory of mind (ToM) and executive functioning (EF) abilities of
124 typically developing preschool children aged 3 to 5 years in relation to whether or
not they had a child-aged sibling (i.e. a child aged 1 to 12 years) at home with whom to
play and converse. On a ToM battery that included tests of false belief, appearancereality (AR) and pretend representation, children who had at least 1 child-aged sibling
scored significantly higher than both only children and those whose only siblings were
infants or adults. The numbers of child-aged siblings in preschoolers’ families positively
predicted their scores on both a ToM battery (4 tasks) and an EF battery (2 tasks), and
these associations remained significant with language ability partialled out. Results of a
hierarchical multiple regression analysis revealed that independent contributions to
individual differences in ToM were made by language ability, EF skill and having a childaged sibling. However, even though some conditions for mediation were met, there was
no statistically reliable evidence that EF skills mediated the advantage of presence of
child-aged siblings for ToM performance. While consistent with the theory that
distinctively childish interaction among siblings accelerates the growth of both ToM and
EF capacities, alternative evidence and alternative theoretical interpretations for the
findings were also considered.
When preschool children develop a theory of mind (ToM), they gain a fundamental
understanding of human psychology that allows them to interpret people’s behaviour in
terms of mental states such as intention, desire and belief. Even more impressively, ToM
includes awareness that people may act on the world in terms of beliefs that are false,
and may be guided by mistaken perceptions, erroneous memories or idiosyncratic
desires and emotions. Consequently, once children acquire a ToM, more effective
communication, interaction and social understanding become possible, and opportunities to engage with siblings and peers in conflict resolution and games of strategy or
pretend are likely to increase (Astington, 2003; Cutting & Dunn, 1999), possibly
accompanied by heightened sensitivity to criticism (Dunn, 1995) or new capacities for
antisocial deception and Machiavellian manipulation (Repacholi, Slaughter, Pritchard, &
Gibbs, 2003).
* Correspondence should be addressed to Professor C.C. Peterson, School of Psychology, University of Queensland, Brisbane,
Queensland 4072, Australia (e-mail: [email protected]).
DOI:10.1348/026151005X70094
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
734 Anna McAlister and Candida C. Peterson
Empirically, the child’s level of ToM understanding is typically measured using
experimental tests of awareness of mental representation involving concepts such as
false belief, appearance-reality (AR), desire reasoning, pretending or thinking (see
Flavell, 1999, for a review of many of these tasks). The most widely used ‘litmus’ test for
ToM is the inferential false belief paradigm. In this procedure, participating children
know the true states of affairs, but must predict the behaviours of protagonists (humans,
puppets or storybook figures) who do not, but instead hold false beliefs about them.
Accurate inference that the protagonist will act upon the faulty belief provides strong
evidence that the child possesses a representational ToM, operationally defined as
awareness of subjective mental states and their mediating role in human behaviour.
Wellman, Cross, and Watson (2001) conducted a meta-analysis of 178 studies of
preschoolers’ performance on false belief tasks like these. The results showed that the
procedure is methodologically robust and reliable, and that, consistently across diverse
sample populations, typically developing children’s performance displays a dramatic
developmental shift between the ages of 2 and 5 years. Wellman et al. found that a
majority of 3-year-olds performed at chance, or below, on false belief tasks whereas, by
age 5, success was so widespread across all task variations as to suggest that
‘understanding of belief, and, relatedly, understanding of mind, exhibit genuine
conceptual change in the preschool years’ (p. 655). Equally importantly, clear individual
differences during this critical acquisition period were revealed, with some
preschoolers consistently succeeding on false belief tests many months ahead of others.
The focus of the present study is on these individual differences in typically
developing children’s rates of ToM development between age 3 and age 5. Several
different explanations have been put forward to account for this variability in
developmental timing. As Slaughter and Repacholi (2003) noted, research striving to
explain individual differences in false belief performance falls into three broad
groupings, namely: (1) studies of early experiences, and background factors in
children’s social-cultural and family environments, that might predict faster or slower
development; (2) studies of neuroanatomical predictors and/or of links between ToM
and basic cognitive skills (e.g. executive functioning); and (3) studies addressing social
outcome measures, such as peer popularity and school adjustment.
Siblings and ToM
Notable among studies of the first type have been those that have explored the
influence of the presence of siblings on ToM development. Sparked in part by
naturalistic longitudinal observations (e.g. Brown, Donelan-McCall, & Dunn, 1996;
Dunn, 1994, 1995) of links between children’s more advanced performance on
laboratory false belief tests and their frequent talk about cognitive and emotional states
during family conversations with siblings and parents, the child’s sibling constellation
has been examined as a possible predictor variable. In one groundbreaking study,
Perner, Ruffman, and Leekam (1994) tested 76 children aged 3 to 5 years on a narrative
false belief test and discovered a linear improvement in performance with increasing
family size. Only children did worse than their peers of the same age who had one
sibling, younger and older siblings were equally beneficial, and children with two or
more siblings significantly outperformed those who had only one. Perner et al.
concluded, ‘the benefit children get from interacting with two siblings rather than
none is worth about as much as 1 year of experience, which at this young age is
considerable’ (p. 1230).
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
735
Subsequently, Ruffman, Perner, Naito, Parkin, and Clements (1998) undertook a
larger study that subsumed the original sample tested by Perner et al. (1994) and found
that the child’s number of older siblings predicted false belief scores over and above
chronological age, but that once age and older siblings were considered, the presence or
numbers of younger siblings exerted no additional benefits. Using a similar method,
Lewis, Freeman, Kryiakidou, Maridaki-Kassotaki, and Berridge (1996) similarly
concluded that older siblings were more beneficial than younger siblings to ToM
development. However, when Jenkins and Astington (1996) controlled statistically for
both chronological age and verbal ability, they found that preschoolers with at least one
younger or older sibling significantly outperformed only children on a false belief
battery, and that family size, but not birth order, was responsible for the sibling effect, so
that ‘[i]t is the number of siblings that the child has that is important for the
development of false belief rather than whether these siblings are older or younger and
how far distant in age they are’ (p. 75).
As Peterson (2000) pointed out, one possible reason for the discrepancies among
these studies’ conclusions, as well as for the disagreement between the results of all
these studies showing a sibling effect and others that have not found it (e.g. Cutting &
Dunn, 1999), may have been the failure of many previous studies to consider the
sibling’s age. Within certain age limits, sibling interaction is conceivably capable of
producing distinctive kinds of social experiences that could help to accelerate a child’s
ToM development. However, outside these age boundaries, sibling interaction may not
be so distinctive or so influential. For example, until they acquire language and social
awareness, infants are probably too young to effectively play or communicate with an
older brother or sister. Thus, a preschooler whose only sibling is a an infant less than 12
months old (an increasingly common situation in middle-class families today, given
contemporary trends for small, widely spaced, sibling constellations) might not have
optimal access to the kinds of sibling interactions that could most effectively stimulate
an awareness of others’ minds, even though observation of parent-infant interaction
might supply some clues. At the other extreme, a sibling who is an adolescent or young
adult is unlikely to play or interact with the preschooler in a way that differs distinctively
from a parent’s. Thus, having a sibling might be most beneficial for ToM growth when
child siblings are the right age to want to share together in pretend play and other
childish games, and are close enough in age to be able to argue, converse, tease, joke,
empathize, compete, cooperate and negotiate with on another on a relatively equal
footing (Piaget, 1965).
Empirically, it is unfortunate that many sibling ToM studies, including those listed
above, have failed to supply full details, or conduct analyses, of siblings’ ages in relation
to these likely optimal interaction boundaries. Thus, the possibility that samples in some
studies contained mostly infant siblings too young to interact meaningfully, and/or many
older siblings who functioned as adults, cannot be ruled out. Discrepancies among
studies’ findings could conceivably be explicable on this basis. (For example, a study of
participants who mostly had infant siblings could produce the result that numbers of
younger siblings are of no ToM benefit, while the opposite conclusion could be reached
if all the younger siblings in a particular study happened to be toddlers or young
preschoolers).
To test this possibility, Peterson (2000) administered a battery of false belief tests to a
sample of 265 Australian children aged 3 to 5 years whose parents had supplied full
information on the ages of all siblings in the participating household. As had been
predicted by the interaction-opportunity hypothesis, children whose only siblings were
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
736 Anna McAlister and Candida C. Peterson
infants under the age of 12 months were found to display no sibling ToM advantage,
performing like only children on false belief tests. The same was true of participants
whose only siblings were teenagers or adults. However, those preschoolers in Peterson’s
sample who had at least one younger or older sibling in the childhood age range scored
significantly higher on false belief tests than both only children and those who only had
infant or adult siblings (dubbed pseudo-singletons). Middle-born children who shared a
household with several child siblings, including those younger and older than
themselves, did best of all, possibly owing to the benefits of the greater variety of
childish interactional opportunities that siblings of mixed ages could supply. Peterson
concluded, ‘[t]he opportunity to play, converse and disagree in distinctively childish
ways with brothers and sisters provides unique insights into the workings of the human
mind’ (p. 435).
While this conclusion seems plausible in terms of interaction opportunity, it is
important to test the replicability of the sibling age boundary effect, and how well it
generalizes to other measures of mentalistic understanding besides false belief (these
were the only ToM tests Peterson (2000) used). As subsidiary aims, these questions will
be addressed in the present study.
Executive functioning, ToM and siblings
Slaughter and Repacholi’s (2003) second type of account for individual differences in
preschoolers’ rates of ToM development centres on differences in cognitive ability,
especially executive functions (EF). Executive capacities are domain-general problemsolving tools. Children with sound EF skills are able to exert self-control and to
strategically inhibit prepotent, but incorrect, responses (Perner & Lang, 1999). They are
also able to think flexibly and planfully about varied kinds of problems while switching
rapidly among different modes of processing information (Carlson & Moses, 2001).
Specific executive skills that are often experimentally measured include forward
planning, mental flexibility, working memory and response inhibition. Preschoolers’
scores on ToM tasks have been shown to correlate with scores on tests of these EF
abilities (Carlson & Moses 2001; Carlson, Moses, & Hix, 1998; Hughes, 1998). For
example, Carlson and Moses found that, even after partialling out children’s language
ability and age, scores on an EF battery of inhibitory control tasks remained strongly
correlated with scores on a false belief battery. Theoretically, some accounts of ToM
development (e.g. Hughes, 1998) postulate that EF abilities are necessary prerequisites
for the basic conceptual understanding of mental states that defines ToM and/or for the
execution of accurate responses to ToM tests. One possibility is that children with
underdeveloped EF skills may not be capable of conceptualizing representational
mental states like false belief. Then again, their implicit conceptual understanding may
fail to produce success on standard tasks owing to the executive demands for memory,
attention, inference and response inhibition that are bound up with the false belief
testing procedure. Alternatively, ToM abilities are sometimes thought of as causally and
developmentally primary to EF skills (e.g. Perner & Lang, 1999), so that an initial
capacity to understand ToM concepts like false belief may underpin the later
development of EF abilities. For example, children may need to reflect on the mind and
understand its links with overt behaviour before they can deliberately employ the
cognitive strategies identified as executive functions.
In yet another intriguing theoretical slant on the relationship between EF and ToM,
Cole and Mitchell (2000) postulated that the presence of siblings might accelerate
children’s development of EF abilities, with these EF advances, in their turn, accelerating
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
737
the development of ToM. As they suggested, ‘[i]t might be that having siblings primarily
promotes executive control’ (p. 281) so that any effect of siblings on ToM performance
is mediated through individual differences in EF.
As Cole and Mitchell noted, the suggestion that siblings could be beneficial to
preschoolers’ EF performance might initially seem unlikely. As they point out, ‘tests of
executive functions often correlate with tests of intelligence’ (p. 280). Yet, Zajonc,
Markus, and Markus (1979) found negative correlations between intelligence test (IQ)
scores and the total number of siblings in the child’s family, and Glass, Neulinger, and
Brim (1974) found that only children and eldest offspring in widely spaced families
(who had grown up as only children for a significant time) scored higher than laterborns on tests of reading ability and scholastic achievement as well as IQ. However, on
plausible theoretical grounds, Cole and Mitchell made the opposite prediction for EF in
preschoolers. Specifically, they argued that young children with siblings might develop
EF skills rapidly through play and sibling interaction that could cultivate EF skills for
remembering, planning strategically and inhibiting prepotent incorrect responses in a
manner not available to the only child. For example, siblings may react more negatively
than parents to one another’s failures to share, plan, attend or control their impulses.
Furthermore, while playing strategic games, negotiating over conflicts, delaying
gratification or taking turns, siblings might observe, approve, teach and copy one
another’s EF skills.
However, when they conducted two experiments that were designed to test
whether the sibling ToM advantage reflected a more basic beneficial effect of siblings
upon EF abilities, Cole and Mitchell (2000) obtained null results. In a sample of
predominantly low-income families, they found mostly non-significant correlations
between preschoolers’ EF scores and their numbers of siblings. Nor was there any
evidence of a sibling ToM advantage, since non-significant correlations emerged in both
experiments between all tasks in their ToM battery and each aspect of sibling numbers
(total, older or younger) that they scored. In fact, the only exception to this
predominant pattern of non-significant findings was a weakly significant correlation
(r ¼ :24) in the second experiment between sibling numbers and one of their four EF
tests. Yet, this same correlation was far from significant in the first experiment (r ¼ :04)
and it likewise fell to non-significance in the second experiment once differences in
children’s verbal abilities were statistically controlled. Cole and Mitchell speculated that
their sample’s socio-economic disadvantage may have contributed to these null results.
In fact, degree of socio-economic disadvantage negatively predicted false belief scores in
their second experiment even after age and verbal ability were partialled out. It makes
some sense that explanatory, mentalistic discourse among siblings, and between siblings
and parents, might be less frequent in lower-income families than in middle-class homes.
Thus, the null results might have reflected a confound between family size and
economic status for the particular groups of children Cole and Mitchell had tested.
Yet, when Hughes and Ensor (2005) studied similar questions among 2-year-olds who
were also predominantly from lower-income families, they did observe a sibling
influence of a different kind. As in Cole and Mitchell’s study, there were no significant
correlations between the toddler’s total number of siblings and his or her scores on ageappropriate batteries of ToM and EF tasks. However, the quality of the children’s
relationships with their siblings (a variable reflecting maternal reports of their frequent
pretend play, frequent affectionate communication, etc.) was a statistically significant
predictor both of toddlers’ ToM performance and of their EF scores. Furthermore, the
link between ToM and sibling relationship quality remained significant even after
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
738 Anna McAlister and Candida C. Peterson
the influences of age, EF, linguistic maturity and parent-child relationship quality were
statistically controlled. Since Hughes and Ensor did not report the ages of their
participants’ siblings, it is unclear to what extent sibling interaction opportunity, as a
function of age, might have contributed to the findings. Infant siblings may have been
too young, and adolescent siblings too old, to engage in the kinds of childish
conversations, pretend play and so on, that would not only produce high
relationship quality scores but also, according to Peterson (2000), promote the rapid
development of ToM.
Aims of the present study
The main aim of the present study was therefore to further examine the extent to which
the presence or absence of child siblings, and sibling numbers and birth order, might be
linked with individual differences in preschooler’s performance on ToM and EF tasks. By
including measures of sibling age, we were able to indirectly test the sibling-interactionopportunity hypothesis (Peterson, 2000); namely, that siblings should be children to
offer optimal ToM benefit. If it is the opportunity for distinctively childish play and
conversation with siblings that is important, then we predict that children with at least
one child sibling (aged 1 to 12 years) will outperform only children and pseudosingletons (i.e. those with only infant or adult siblings) on batteries of standard ToM and
EF tasks. In addition, by examining a larger sample of children than in Cole and
Mitchell’s (2000) study, as well as a group more economically advantaged than theirs,
and by including additional ToM tests of AR and pretend representation (PR) as well as
false belief, subsidiary aims of the study included more extensive testing of two plausible
hypotheses put forward by Cole and Mitchell, namely: (1) that a more affluent sample of
preschoolers than in Cole and Mitchell’s study would reveal clearer links between the
three target variables (child siblings, EF and ToM), and (2) if so, that evidence of superior
ToM performance by participants with child siblings would be mediated through a more
basic child-sibling influence upon EF.
Method
Participants
After the exclusion of 21 children (3 for reports of a suspected diagnosis of
developmental delay, 3 for a parental report of a primary family language other than
English, 6 owing to parents’ failure to specify the sibling composition of their family and
9 for failing a control question on a false belief test; see below), the sample consisted of
124 Australian children (71 boys and 53 girls) who ranged in age from 3; 3–5; 9 with a
mean of 4; 3. The children were attending preschools that served predominantly middleclass neighbourhoods, all had English as a sole or primary language and none had a
known or suspected serious intellectual, developmental or sensory impairment. In
addition to predominantly middle-class residence, their parents were well educated
(averaging ‘some university study’; see Table 2).
In terms of sibling composition, 18 children in the sample (14%) were only children
who had no brothers or sisters of any age. An additional 5 children resembled only
children in terms of their family conversational environment (Peterson, 2000) in that
their sole sibling was either a very young infant (2 children) or an adult (3 children). For
the statistical analyses, these groups (singletons plus pseudo-singletons) were merged
together as Group 1, defined as participants with no child siblings. They comprised 19%
of the sample. Group 2 consisted of 49 children (40% of the sample) who had at least
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
739
one older child sibling (under age 13) but no younger child siblings. Of these, 40
children were the youngest in their families while an additional 9 also had an infant
sibling under 12 months of age. Group 3 consisted of 21 children (17% of the sample)
who had both older and younger child siblings and Group 4 consisted of 31 children
(25% of the sample) who had younger child sibling(s) but no older ones.
Tasks and scoring
The children completed measures of ToM (four tasks) and EF (two tasks) plus a language
measure. These tests are described below.
Theory of mind (ToM)
The ToM battery was chosen to include widely used tests of representational
understanding that, like those used by Hughes and Ensor (2005) varied sufficiently in
difficulty level to be able to pick up early signs of ToM awareness while also emphasizing
the standard ‘litmus’ ToM measures like false belief. The ToM tests were of three
different kinds; namely, false belief, AR and pretend representation. All three of these
were alike in requiring children to understand mental representation together with the
fact that people’s inner psychological states can misrepresent objective reality.
Specifically, to pass the false belief tests, children had to predict actions of protagonists
whose beliefs misrepresented the real locations of objects or contents of containers. To
pass AR, they needed to understand that an object’s perceptual appearance can
misrepresent its true nature and function, and to pass the pretend tasks, they needed to
understand the distinction between an object’s real properties and its alternative
representational identity during pretending.
False belief. There were two false belief tests: (a) a standard unseen displacement
task and (b) a standard misleading container task. For (a), we used Baron-Cohen, Leslie,
and Frith’s (1985) ‘Sally-Ann’ task, presented and scored exactly as described by the
original authors except that a boy replaced the second girl doll (Peterson & Siegal,
1999). In brief, each trial began with the girl doll placing a marble in a covered basket
and going away. The boy doll then moved the marble either to a covered box (Trial 1) or
to the tester’s pocket (Trial 2). The girl returned wanting her marble. Children were
asked the test question, ‘Where will the girl look first for her marble?’ and then two
control questions (‘Where is the marble really?’ and ‘Where did the girl put the marble in
the beginning?’). When children passed both control questions on both trials, it was
clear that they had enough language to understand the tasks and test questions and that
they had attended to, and remembered, the critical information on which a valid test
question response could be based. Thus, we required that all children pass these control
questions in order to be included in our sample. (In the rare instance where a child we
had begun to test failed one of these four control questions, he or she was not tested any
further and was not included in our sample or analyses). Each correct test question
response earned 1 point to a maximum total of 2 for the task.
The misleading container false belief test was borrowed from Gopnik and Astington
(1988) and involved a familiar Band-Aid box that actually held pencils. The child was
initially shown the closed box and was asked what was in it. After naming the expected
contents (which all did), the pencils were displayed and the box was resealed. The
control question (‘What is really in the box’) was given, followed by a test question
about other’s belief: ‘[name of class mate] is coming next. He [she] hasn’t seen inside
this box before. When I show it to him [her] all closed up like this, what will he [she]
say is in it?’ Next was a representational change test question about own false belief:
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
740 Anna McAlister and Candida C. Peterson
‘When I first showed you this box, before you looked inside, what did you think was in
it?’ followed by a control question: ‘What is really in it?’ All children passed both control
questions. Correct test question responses (Band-Aids) earned 1 point to a possible total
of 2 for the task.
Children’s scores on the two false belief tasks were found to be statistically
significantly correlated, rð122Þ ¼ :50, p , :001. Consequently, a total false belief (TFB)
score was created by summing scores on the changed location and container tasks. This
total could (and did) range from 0–4.
Appearance-reality (AR). This test was based on Flavell’s (1986) procedure. There
were two trials. The first began with the display of a realistic wax apple that was really a
candle. On being asked what it was, all children said ‘apple’. Then the tester
demonstrated the real identity by lighting the candle and asked the two test questions in
counterbalanced order: ‘What is this really and truly?’(reality test) and ‘When you look at
this with your eyes right now, what does it look like, [pause] does it look like a [candle]
or like [an apple]?’ (appearance test). Accurate understanding of the appearance–reality
distinction requires that the child distinguish the two identities, so that a child who gets
only one test question right (typically by giving the same answer, e.g. apple, to both test
questions) has demonstrated no greater understanding one who gets both wrong.
Indeed, the latter child (who has typically distinguished appearance from real identity
while transposing their labels) could be argued to have somewhat greater understanding
than the former. Consequently, in order to pass, children had to answer both test
questions correctly and, if so, earned 1 point. Other response patterns earned 0.
A second AR trial using a realistic-looking cloth flower that was actually a ballpoint pen
was conducted using identical procedures, questions and scoring to the apple-candle
test.
Scores on the apple-candle and flower-pen versions were significantly correlated,
rð122Þ ¼ :34, p , :01. Consequently, a total AR score ranging from 0–2 was created by
summing scores on both of them.
Pretend representation (PR). The third component of the ToM battery assessed
children’s awareness of the representational nature of pretending and of the contrast
between an object’s real, physical identity and its mental representation as something else
during pretence. There were two trials. Each began with the child being shown a familiar,
real object (banana or potato, counterbalanced) and being asked to pretend it was,
respectively, a phone or a bar of soap. After minimal prompting (e.g. ‘Your phone’s
ringing’), all children successfully engaged in actions that unequivocally showed they
understood and accepted the pretend stipulation (e.g. by holding the banana to their ear,
or rubbing the potato over both hands). After a short period of pretend play, the tester said
‘Okay, we’ve finished our game now. We have stopped pretending’. Then, the tester
pointed at the banana or potato and asked two test questions: (1) ‘What is this really?’ and
(2) ‘What did we pretend this was?’ Two additional objects were then placed on the table
alongside the banana or potato. One was always a real version of the pretend item (e.g. a
real mobile phone) and the other was an irrelevant decoy (e.g. a tennis ball). The last two
test questions for each trial were: (3) ‘Which one did we pretend was a [telephone/soap]?’
and (4) ‘Which one is really a [phone/soap]? Thus, for each PR trial, there were four test
questions, using different syntax, two about the real, and two about the apparent, identity
of the object. For each correct test question response, 1 point was awarded, to a
maximum of 4 for each task. Scores on the two tasks were statistically significantly
correlated, rð122Þ ¼ :31, p , :01, so children’s scores on the two tasks were summed,
then halved to give a total out of 4, similar to the false belief task.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
741
Total ToM score
The children’s total FB scores were significantly correlated with their total AR scores,
rð122Þ ¼ :29, p , :01, and with their total PR scores, rð122Þ ¼ :43, p , :01. The PR
total was also significantly correlated with the AR total, rð122Þ ¼ :30, p , :01.
Consequently, given these significant associations and in order to create a total ToM
score that would given equal weighting to each of these three frequently assessed
aspects of representational understanding, a proportion score for each component was
first computed by dividing the FB and PR totals by 4, and the AR total by 2. These
decimal totals were then averaged into a ToM composite for each child that could range
from 0–1. Cronbach’s test was run on this ToM composite and produced an alpha
reliability coefficient of .65, indicating acceptable internal consistency for the ToM
composite as a summary score.
Executive functioning (EF)
Two tests of EF were borrowed from Cole and Mitchell (2000). The first was a route
navigation (RN) test assessing EF capacities for forward planning, attention to relevant
features, resistance to distraction, and response inhibition. The task was presented and
scored exactly as described by Cole and Mitchell. There were four trials. On each trial,
the child was shown a symmetrical maze printed on coloured card with a picture of a
cartoon character at the entrance (middle-bottom) and a picture of a desired goal (e.g. a
gift-wrapped present) at the middle-top (exit). The paths on the right and left side of
each maze were both of equal length and both had 10 bends. One path (left or right,
counterbalanced over trials) was unobstructed, while the other was blocked at three
points by brightly coloured stickers. Children were told that the route with the stickers
was the ‘hard’ way for the character to go because each sticker represented an obstacle
that had to be climbed over. The route without stickers was described as the ‘easy’ way
because it had no blockages. Children were then asked to draw with their finger the easy
way for the character to go to reach the goal. On the first trial only, corrective feedback
was given to any child who made an error, in which case the trial was repeated. After
this, without correction, children earned 1 point for tracing the unobstructed path,
otherwise 0, to a maximum possible score of 4 for the task. This score was then
converted to a proportion of trials correct by dividing by 4.
The second EF test was Cole and Mitchell’s (2000) adaptation of Reed, Pien, and
Rothbart’s (1984) ‘resisting instructions’ (RI) test. We followed Cole and Mitchell’s
procedure except that a bear and lion replaced their squirrel and badger for the sake of
familiarity to local children. This task assesses EF skills for response inhibition, task
switching and adherence to directions. In brief, over 12 trials, the bear and lion puppets
took turns asking the child to perform various movements (e.g. ‘touch your nose’). The
hand on which the tester wore bear and lion puppets was counterbalanced across
participants. Before the test began, children were told that the lion was ‘naughty’ and
should not be obeyed. Instead, they should keep their hands flat on the table whenever
the lion instructed them to do something. However, the bear was ‘good’ so they should
always do exactly what the bear told them to. Initially, and again at the start of every
fourth trial, the tester asked the child to point to the puppet who should be obeyed and
to the one who should be ignored. In randomized order, the bear gave the command on
6 trials, and the lion on the remaining 6 trials. The lion’s trials were test trials and the
bear’s were filler trials and were not scored. Children earned 1 point for each trial in
which they ignored the lion and kept their hands perfectly still. Thus, the maximum
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
742 Anna McAlister and Candida C. Peterson
possible total was 6 for the task as a whole. To yield a total similar to the RN, this score
was converted to a proportion by dividing by 6.
Scores on the RN and RI tasks were statistically significantly correlated, rð122Þ ¼ :33,
p , :01, so children’s scores on the two tasks were averaged to yield a composite EF
total score.
Language ability
The Peabody Picture Vocabulary Test (PPVT; Dunn & Dunn, 1981) was administered
and scored according to standard instructions to give three estimates for each child.
These were: (1) the raw score (a measure of receptive vocabulary size), (2) a verbal
mental age estimate (VMA; or language maturity relative to children in Dunn & Dunn’s
1981 American normative sample) and (3) a standard score (a deviation-type norm, or
Verbal IQ equivalent).
Parent education
The parent who supplied information on sibling composition, home language and
disability diagnosis also completed a brief scale of educational attainment for self and
spouse by selecting, in response to the question, ‘What is the highest level of education
presently achieved?’, a response option from the following: (a) ‘Year 10 only’ (the
minimum legal school leaving age), scored 1; (b) ‘high school or technical school
completion’ (12 years of schooling), scored 2; (c) ‘some university studies but no
degree’, scored 3; (d) ‘university Bachelor’s degree completed’, scored 4; (e) ‘university
Masters or PhD degree completed’, scored 5. Father’s and mother’s scores were
averaged, for the 119 intact families in the sample, to produce a parent mean score that
could range from 1–5. For the five single mothers without a co-resident partner, the
mother’s score was used without averaging.
Testing procedure
Each child was tested individually in a quiet area of the preschool in two sessions,
separated by about one week. Each session include a single member of the FB, AR, PR
and EF task pairs. The specific task of each type that was given first or second was
counterbalanced. The PPVT was always given in the second testing session.
Results
Descriptive information and preliminary analyses
Table 1 shows the children’s mean scores on the ToM and EF measures. In order to
confirm that children in this sample were developing on the typical timetable that has
Table 1. Children’s performance on ToM and EF tasks
ToM battery
Possible score
Mean score (SD)
Range
EF battery
False belief
(FB)
Pretend representation (PR)
Appearance
reality (AR)
Route
navigation (RN)
Resisting
instructions (RI)
4
2.54 (1.44)
0–4
4
3.52 (0.64)
1–4
2
.78 (0.82)
0–2
4
3.19 (1.05)
0–4
6
3.25 (2.69)
0–6
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
743
been established in much previous research (see Wellman et al., 2001, for a review), age
trends for our sample’s ToM scores were examined. As expected based on these earlier
studies, ToM composite scores for the children in this sample were significantly
correlated with age, rð122Þ ¼ :31, p , :01, two-tailed. Furthermore, when the sample
was subdivided into three age groupings of (1) 3-year-olds (N ¼ 42; mean age ¼ 3.72),
(2) young fours (N ¼ 43; mean age ¼ 4.23) and (3) older preschoolers (N ¼ 39; mean
age ¼ 4.87), a significant age difference in ToM scores emerged, Fð2; 127Þ ¼ 4:72,
p , :05. To locate the basis for this difference, single-df planned comparisons were used
(Keppel, 1982). The 3-year-olds, with a mean proportion of 0.67 ToM test trials correct,
were found to be significantly less advanced than the young 4-year-olds who, with a
mean of 0.78, Fð1; 122Þ ¼ 8:11, p , :01, did as well as the older preschoolers, whose
mean was .77, Fð1; 122Þ , 1, ns. There were no statistically significant differences
among the three age groups in level of parent education, F , 1, child gender
distribution, x2 ð2; 124Þ ¼ 2:33, ns, or PPVT standard scores, F , 1. Thus, the age effect
was an uncomplicated one, and it supported previous findings of a rapid surge in ToM
development after the fourth birthday (Wellman et al., 2001). Consequently, in contrast
to Cole and Mitchell’s (2000) socio-economically disadvantaged sample, there was no
reason to anticipate lack of sibling influences on ToM for children in the present sample
owing to delayed mentalistic understanding.
As a further preliminary analysis, we examined the comparability of the four sibling
constellation groups on variables that could potentially confound the assessment of
sibling effects on ToM and EF. Table 2 gives shows the groups’ mean scores on these
variables. Results of the preliminary analyses revealed that there were no statistically
significant differences among the sibling constellation groups in age, Fð3; 120Þ ¼ :53,
ns, in gender balance, x2 ð3; 124Þ ¼ 6:27, p ¼ :10, in mean levels of parents’ education,
Fð3; 120Þ ¼ 1:49, ns, or in language maturity (VMA), Fð1; 122Þ , 1, ns. Thus, neither
the contrast between only children and those with a child sibling, nor the participant’s
status as a youngest, middle or eldest child sibling, displayed any consistent associations
with any of these other potential predictors of ToM or EF, consequently simplifying the
main statistical analyses, and indicating that any differences we observe in ToM or EF
abilities as a function of sibling status will not be explicable as an indirect consequences
of these other variables.
Child siblings and ToM development
To test the set of hypotheses relating to the influence of child siblings on individual
differences in rates of ToM development, an analytic set of single-df planned
comparisons was conducted, in the manner recommended by Keppel (1982) ‘directly
on [the] data without reference to the significance or non-significance of the omnibus
F test’ (p. 106) and ‘instead of the overall F test’ (p. 104, emphasis in original text). The
first hypothesis, derived from Peterson (2000), was that the presence of at least one
child sibling in the family would be linked with significantly more advanced ToM than
for Group 1 children without siblings at home. This was tested with a planned
comparison contrasting Group 1’s ToM composite score with that of the combined
remainder of the sample. The result was statistically significant, Fð1; 120Þ ¼ 6:08,
p , :025, indicating that the children without child siblings in Group 1 were
significantly less advanced in ToM development than those who had at least one brother
or sister aged 1 to 12 years. Next, to test the hypothesis derived from Ruffman et al.
(1998) that older child siblings might facilitate ToM growth whereas younger siblings
would not, a planned comparison was conducted between Groups 2 and 3 combined
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
744 Anna McAlister and Candida C. Peterson
(both of whom had older child siblings) versus Group 1 plus Group 4 (without older
child siblings). The result was not statistically significant, Fð1; 120Þ ¼ 2:72, ns. A
different way of testing the same hypothesis, using the strategy employed by Jenkins and
Astington (1996), is to compare children who have older siblings only with those whose
only siblings are younger, excluding middle-borns. This planned comparison, between
Group 2 and Group 4, was similarly non-significant, F , 1, ns. Finally, in order to test the
hypothesis from Perner et al. (1994) that children with two or more child siblings
develop ToM faster than those with only one child sibling, a planned comparison was
conducted between Group 3 (who had a mean of 2.10 child siblings) and Groups 2 and
4 combined (who had a mean of 1.05 child siblings). This comparison was likewise nonsignificant, F , 1, ns.
In other words, children in the present sample who had no child siblings scored
significantly below those who had a child sibling on our composite ToM measure. These
findings support Peterson’s (2000) for children from a similar population. In particular,
using a different ToM battery that included additional AR and PR tasks, our results
replicated Peterson’s findings (1) that better false belief performance was associated
with having at least one child sibling, and (2) that the sibling’s being older, rather than
younger, than the participant was not additionally beneficial.
Siblings and EF
Table 2 shows the mean EF composite scores for children in each of the sibling
constellation groups. The possibility that having a child sibling would influence EF
performance was tested using single-df planned comparisons (Keppel, 1982). For the
first comparison, participants without child siblings (Group 1) were contrasted with the
combined remainder of the sample (Groups 2, 3 and 4) who had at least one child sibling
aged 1 to 12 years. This contrast was statistically significant, Fð1; 120Þ ¼ 4:12, p , :05.
In other words, participants with a child sibling did better on EF tasks than only children
plus those whose only sibling was an infant or an adult.
The second planned comparison evaluated influences of the presence of an older
sibling by contrasting the EF scores of Groups 1 and 4 with those for Groups 2 and 3
Table 2. Performance by children in different sibling status groups on theory of mind, executive
functioning and language measures, along with other sample characteristics
Participant’s sibling situation
Group
Group size
Mean age (SD)
Age range
Mean parent
education (SD)
Gender balance
(male: female)
Mean VMA (SD)
Mean ToM Total (SD)
Mean EF Total (SD)
1. No child 2. Has older child
siblings
sibling(s) only
3. Has younger and
older child siblings
4. Has younger
child sibling(s) only
N ¼ 23
4.14 (0.59)
3;3–5;2
3.16 (1.01)
N ¼ 49
4.26 (0.59)
3;3–5;9
3.12 (1.17)
N ¼ 21
4.32 (0.48)
3;6–5;5
3.40 (1.08)
N ¼ 31
4.30 (0.48)
3;3–5;4
3.65 (1.18)
13:10
25:24
17:4
16:15
4.47 (0.97)
0.66 (0.19)
0.65 (0.27)
4.41 (1.12)
0.75 (0.17)
0.73 (0.27)
4.63 (0.96)
0.78 (0.21)
0.85 (0.18)
4.80 (1.51)
0.75 (0.20)
0.74 (0.26)
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
745
(participants with at least one older child sibling). The result approached, but did not
achieve, statistical significance, Fð1; 120Þ ¼ 3:50, p ¼ :06. The alternative way of
testing benefits of older versus younger siblings (by contrasting Group 2 with Group 4)
was likewise non-significant, F , 1, ns. In other words, for the children in this sample,
higher EF scores were linked with the presence of a child sibling, whether younger or
older, rather than being confined specifically to the presence of older siblings who
might serve tutoring, or other pseudo-adult, roles.
Correlations of ToM and EF with sibling variables
Table 3 shows the Pearson univariate correlations that emerged among key variables in
this study. In line with much previous research, there were significant correlations of
children’s ToM and EF scores with their chronological age and verbal ability (PPVT raw
scores). With verbal ability partialled out, however, the correlation between ToM and EF
remained statistically significant, rð121Þ ¼ :37, p , :001. Within this relatively welleducated sample, parent education was not significantly related to any variable we
measured.
Table 3. Univariate correlations among children’s scores on theory of mind, executive functioning,
language measures and background variables
Variable:
CA
#Sibs
ChSib
VMA
ToM
EF
Chronological age (CA)
Number of child siblings (#Sibs)
Presence of child sibling (ChSib)
Verbal mental age (VMA)
ToM total (ToM)
EF total (EF)
1.00
.04
1.00
.11
.72**
1.00
.63**
2 .03
2 .10
1.00
.31**
.20*
.22**
.52**
1.00
.54**
.20*
.17
.46**
.48**
1.00
Note. * denotes significant at p , :05 two-tailed, ** at p , :01, two-tailed
In line with the ANOVA results, ToM scores were positively correlated with the
presence of at least one child sibling in the participant’s household (dummy coded as
0 or 1, as recommended by Tabachnick & Fidell, 1989), and this correlation remained
significant even after verbal ability was partialled out, rð121Þ ¼ :29, p , :01. Similarly,
the child’s total number of child-aged siblings was correlated with ToM, and remained
significant after partialling out verbal ability, rð121Þ ¼ :23, p , :02. Like ToM, EF
scores were significantly correlated with the preschooler’s number of child siblings,
rð121Þ ¼ :20, p , :03, and this correlation remained significant with verbal ability
partialled out, rð121Þ ¼ :23, p , :02. EF scores were also marginally correlated with
the presence of at least one child sibling in the participant’s household ( p ¼ :07) and
this correlation rose to statistical significance after verbal ability was partialled out,
rð121Þ ¼ :23, p , :02. Overall, then, in contrast to Cole and Mitchell’s findings for
lower-income children of similar age to the present sample, but partially in line with
Hughes’ and Ensor’s (2005) findings for younger lower-income children, we found
that individual differences among children in our sample in both ToM and EF
performance were linked with their access at home to opportunities for play and
conversation with child siblings who were old enough to be genuine interactional
partners, yet not too old to converse or play in distinctively childish ways.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
746 Anna McAlister and Candida C. Peterson
Furthermore, correlations of the child sibling variables with ToM and EF remained
statistically significant even after differences in children’s language scores were
statistically controlled.
Hierarchical multiple regression analyses
In order to examine more closely the separate contributions of child siblings to
individual differences in children’s ToM and EF scores, even after accounting for the
mutual dependency of both EF and ToM on age and language, we conducted a number
of hierarchical multiple regression analyses.
For the first set of analyses, ToM scores served as the dependent variable. Age and
PPVT raw scores were entered together as control variables at Step 1 in each analysis.
For the first analysis, the variable ‘presence of a child sibling’ (dummy coded: 1 ¼ at
least one child-aged sibling versus 0 ¼ none, as recommended by Tabachnick & Fidell,
(1989)) was entered at Step 2. At the end of Step 1, with only the control variables in the
equation, there was a significant effect, Mult. R ¼ :42, R 2 ¼ :18, Adj. R 2 ¼ :17,
p , :001. At the end of Step 2, with the presence of a child sibling included, there was a
statistically significant increment in the amount of ToM variability that was predicted,
FðchangeÞ ¼ 8:66, p , :01. In other words, information on whether or not the
preschooler had access at home to a child-aged sibling improved the prediction of
differences in ToM scores beyond that afforded by chronological age and verbal ability
combined. With EF scores entered at the third step, a further significant increment
emerged, FðchangeÞ ¼ 11:52, p , :01. In other words, information on differences in the
children’s EF abilities further improved the prediction of differences ToM, beyond that
afforded by age, language and child siblings. At the end of this final step, with all the
predictor variables included, the full model was statistically significant, Mult. R ¼ :56,
R 2 ¼ :32, Adj. R 2 ¼ :29, p , :01, and three independent variables were found to have
statistically significant beta weights in the final equation. These were verbal ability
(b ¼ 0.32, t ¼ 3:33, p , :01), the presence of a child-aged sibling (b ¼ 0:22, t ¼ 2:60,
p , :02), and EF scores (b ¼ 0:33, t ¼ 3:39, p , :01). Thus, language, having a child
sibling, and EF performance each made significant independent contributions to
predicting individual differences in children’s ToM understanding.
To test whether order of entry of the EF and sibling variables made a difference to the
outcome, we conducted a similar hierarchical regression with the ordering of Steps 2
and 3 transposed, so that EF scores were entered immediately after the control variables
(age and language), and the presence of a child sibling was entered at the last step. The
result was very similar to that with the original order of entry. The presence of a child
sibling produced a statistically significant increment at the final step, even after
variability due to EF, age and language had been taken into account, FðchangeÞ ¼ 6:74,
p , :02, and the beta weights at the end of the final step were the same as in the initial
regression equation, with verbal ability, EF and child siblings each making significant
independent contributions. A third set of hierarchical multiple regressions predicting
ToM with numbers of child siblings in place of the previous child sibling variable
(presence/absence) likewise produced similar results except that, in this case, the
increment was statistically significant when the number of child siblings was introduced
at the second step, FðchangeÞ ¼ 5:12, p ¼ :03, but only marginally significant when
number of child siblings was entered after EF at the third step, FðchangeÞ ¼ 2:57,
p ¼ :11. In other words, comparing all these regression models, variability in children’s
ToM performance was clearly predicted, over and above language and age, by EF skill
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
747
and by the presence of a child sibling, and presence of at least one a child sibling was a
more important independent predictor than the total number of child siblings in the
household.
A second set of hierarchical multiple regression analyses was conducted with EF
scores as the dependent variable (DV) and ToM and numbers of child siblings as
predictors. As for the ToM regressions, child age and PPVT raw scores were entered
together as control variables at Step 1. The equation was significant at the end of this
step, Mult. R ¼ :56, R 2 ¼ :31, Adj. R 2 ¼ :30, p , :001. When number of child siblings
was entered at Step 2, there was a statistically significant increment in the variability in
EF scores that was predicted, FðchangeÞ ¼ 5:17, p , :05, and the beta weight for
number of child siblings was separately significant, b ¼ 0:18, t ¼ 2:27, p , :05. ToM
scores were entered at the third step and a further significant increment emerged,
FðchangeÞ ¼ 10:76, p , :01. At the end of this final step, with all the predictor variables
included, the full model was statistically significant, Mult. R ¼ :64, R 2 ¼ :40, Adj.
R 2 ¼ :38, p , :001, and two independent variables were found to have statistically
significant beta weights in the final equation; namely, chronological age (b ¼ 0:40,
t ¼ 4:48, p , :001), and ToM scores (b ¼ 0:28, t ¼ 3:28, p , :01). A similar result
emerged when ToM scores were entered at Step 2 and number of child siblings was
entered last. Thus, the number of child siblings, while significant beyond age, did not
independently contribute to predicting variability in children’s EF performance once
both age and ToM scores were considered, suggesting that number of child siblings
overlapped with ToM in predicting EF skill.
Test of EF as a mediator
A final set of regression analyses was conducted in order to test Cole and Mitchell’s
(2000) hypothesis that the effect of child siblings upon individual differences in
preschoolers’ ToM performance might be mediated through a more basic effect of
siblings upon EF, which might then drive the ToM advantage. We followed the Baron and
Kenny’s (1986) approach to testing mediation, using the regression design that was
recommended by Holmbeck (1997). In the model we tested, with effects of the control
variable (chronological age) statistically controlled, as recommended by Holmbeck, and
with the number of child siblings as the independent variable (IV), EF performance as
the mediator (M), and ToM scores as the DV, three out of required four conditions for
mediation were met, namely: (1) the participant’s number of child-aged siblings (IV)
significantly predicted the mediator (EF), p , :05, (2) the participant’s number of childaged siblings (IV) significantly predicted the DV (ToM), p , :05, and (3) EF performance
(M) significantly predicted ToM (the DV) when it was entered together with number of
child siblings in a standard regression equation, p , :01. However, the fourth condition
for mediation was not met; namely, that there should be a significant reduction in the
ToM variability that was predicted by the IV (child siblings) after inclusion of the
mediator (EF). The results of a Sobel test indicated that this reduction was not
statistically significant, Sobel ¼ 1.83, p ¼ :07. Furthermore, a Goodman I test
(recommended by Baron and Kenny, 1986, as an alternative to the Sobel test) showed
the same thing: Goodman I ¼ 1:78, p ¼ :07. In other words, the data failed to
significantly support mediation. Even though the variables were intercorrelated, there
was no statistically reliable evidence that the advantage of numbers of child siblings for
ToM was channelled through EF.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
748 Anna McAlister and Candida C. Peterson
Discussion
The present results for ToM performance replicated Peterson’s (2000) finding of more
advanced false belief understanding among preschoolers with at least one child-aged
sibling at home with whom to play and converse than for only children and pseudosingletons who had no child siblings. Our results also importantly extended upon these
earlier findings by showing that the child-aged sibling advantage applied also to
preschoolers’ EF performance (in line with Cole and Mitchell’s 2000 prediction but not
their empirical findings) as well as to a broader range of ToM capacities for a
representational understanding of pretence and perceptual appearance, in addition to
false belief. Furthermore, results of hierarchical multiple regression analyses showed
that sibling interaction opportunity (operationalized as having at least one child-aged
sibling at home with whom to play and converse) explained additional individual
variability in children’s ToM scores beyond that accounted for by age and linguistic
maturity. EF scores also predicted ToM performance even after the effects of language
skill and age were considered, and when all four of these predictors were considered
together at the final steps of hierarchical regressions, significant independent
contributions to ToM variability were found to arise out of three of them, namely: (1)
sibling interaction opportunity, (2) language ability and (3) EF skill.
Hierarchical regression analyses were also conducted with EF scores as the
dependent variable. Results of these indicated that numbers of child siblings predicted
variability in preschoolers’ EF performance beyond that explained by chronological age
and language ability alone. However, child siblings did not independently contribute to
predicting variability in EF performance once both age and ToM were considered.
A number of causal accounts for results like these are conceivable and, on the basis
of a cross-sectional correlational design like ours, no single developmental or causal
pathway can be uniquely identified. Longitudinal investigation of the same variables
would be useful for this purpose. Nevertheless, purely speculatively, it is possible that
the nature of the social interaction that goes on in a preschooler’s family when a child
sibling is present differs qualitatively from that available to a singleton or pseudosingleton whose only social companions are parents or other adults. The added social
experiences available in the former instance could conceivably underpin the childsibling advantages for ToM and EF performance. For example, as Cole and Mitchell
(2000) suggested, a preschooler’s EF skills for processing information strategically and
flexibly while inhibiting impulsive responses could be enhanced through experiences
with siblings of playing competitive, cooperative and strategic games, by older siblings’
modelling or coaching of these skills, or via sibling disapproval when EF skills are not
manifested. Similarly, as Hughes and Ensor (2005) suggested, frequent opportunities to
play imaginatively with siblings, to discuss thoughts and feelings and to resolve sibling
conflicts could stimulate not only the social awareness that is bound up with ToM but
also the cognitive flexibility, inhibitory control, rule-consistency and problem-solving
skills that are assessed by many EF measures.
Alternatively, it is conceivable that other factors, and a different chain of influence,
might account for the correlations we observed. For example, parents who plan their
families with close sibling spacing because they want to make sure each of their children
has companionship from a similar-aged brother or sister might also differ in other ways
from parents with different family goals, or those who do not actively plan child-bearing.
Associated differences in parental philosophies or child rearing strategies might then
indirectly explain the correlations between child siblings and both EF and
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
749
ToM performance. It is also likely that parents’ styles of talking with their offspring about
thoughts and feelings may have a bearing on how readily the children develop ToM
(Brown & Dunn, 1992; Dunn, 1994; Peterson & Slaughter, 2003). The present results do
not inform directly on how variations in parents’ conversational approaches might
relate to their families’ sibling constellations and it would be useful for future research to
study this. Interestingly, however, Hughes and Ensor (2005) did measure parenting
quality in terms of variables like parental talkativeness and conversational responsiveness, and found that it was a less important predictor of 2-year-olds’ ToM performance
than was positive sibling relationship quality.
Cole and Mitchell (2000) postulated that the effects of a sibling on children’s ToM
performance might be mediated through a more basic effect of siblings on children’s EF
capacities for planning, flexibility and inhibitory control. We tested this mediation
hypothesis. However, despite significant associations among key variables, there was no
statistically reliable evidence of this mediation effect. Within the limits of our sample and
the particular EF and ToM tests we used, the influence of child siblings on ToM did not
appear to be fully channelled through executive cognition. Furthermore, our data
hinted at subtle differences in the manner in which the sibling interaction opportunity
variable was linked with individual differences in EF performance, in contrast to ToM
performance, for these preschoolers. While both of the child sibling variables we
identified, namely (a) the presence of at least one child-aged sibling, and (b) the total
number of child siblings in the family, were linked with ToM performance, links with EF
scores showed a slightly different pattern. The number of child siblings correlated with
EF scores both before and after verbal ability was partialled out, whereas the correlation
with the presence of a child-aged sibling only achieved statistical significance once
verbal ability was statistically controlled.
If having more than just one child sibling turns out to be more important for EF
performance than for ToM (a possibility that requires further empirical investigation),
this could have implications for theories of how child siblings influence preschoolers’
social and cognitive development. While speculative, one possibility is that the total
number of child playmates relates linearly to advanced EF skills because group
interaction with several siblings at once is of special value in cultivating EF. For example,
teamwork may be critical, as when a group of siblings engages in competitive or
strategic games. For ToM, on the other hand, dyadic conversation and pretend play with
just one sibling may also supply developmentally useful input, assisting the child’s
understanding both of other people’s roles and perspectives and of the subjectivity of
inner mental states of believing or pretending. Further research that directly addresses
these questions using observations of family interaction would be valuable, and
additional empirical evidence is clearly needed before such possibilities can be anything
other than highly speculative.
Also in need of further research is the question of how EF and ToM development
interconnect with one another over the course of development. Hughes and Ensor’s
(2005) findings provide important indications of correlations that pre-date the
preschool period, as well as of the importance of high quality interaction among siblings
for both EF and ToM capacities emerging during toddlerhood, and these patterns are
largely consistent with present findings. However, the accumulated empirical evidence
to date still does not permit a definitive choice among existing theoretical models for
how EF relates to ToM. These, as Perner and Lang (1999) noted, fall into three broad
groupings. Some theories postulate that EF abilities are prerequisites for the
development of ToM. Others view ToM as a prerequisite that is needed before EF can
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
750 Anna McAlister and Candida C. Peterson
develop, whereas theories of a third kind postulate that both types of skill may develop
simultaneously, either through coordinated maturation of brain structures and/or
through shared dependency on similar socializing experiences and supporting skills
(e.g. language).
After reviewing the empirical evidence then available, Perner and Lang (1999)
concluded that each of these theoretical possibilities was tenable, and the present
findings continue to align themselves with the same open-ended causal picture. Yet, our
results do make an important addition to the picture by demonstrating that the social
influence of an available child playmate in the family is beneficial to EF performance, as
well as well as to ToM, during the preschool period. It is to be hoped that future
research, ideally incorporating both longitudinal and naturalistic-observational
assessment, can further refine the understanding both of the specific manner in
which interaction among child siblings may influence ToM and EF performance, and
vice versa. Future research, ideally incorporating a longitudinal design, is also needed to
clarify the developmental paths of change and interdependency among these core
cognitive and social-cognitive abilities.
References
Astington, J. A. (2003). Sometimes necessary, never sufficient: False belief understanding and
social competence. In B. Repacholi & V. Slaughter (Eds.), Individual differences in theory of
mind ( pp. 13–38). New York: Psychology Press.
Baron, R., & Kenny, D. (1986). The moderator-mediator distinction in social psychological
research. Journal of Personality and Social Psychology, 51, 1173–1182.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a theory of mind.
Cognition, 21, 37–46.
Brown, J., Donelan-McCall, N., & Dunn, J. (1996). Why talk about mental states? The significance
of children’s conversations with friends, siblings, and mothers. Child Development, 67,
836–849.
Brown, J., & Dunn, J. (1992). Talk with your mother or your sibling: Developmental changes in
early family conversations about feelings. Child Development, 63, 336–349.
Carlson, S. M., & Moses, L. J. (2001). Individual differences in inhibitory control and children’s
theory of mind. Child Development, 72, 1032–1053.
Carlson, S. M., Moses, L. J., & Hix, H. R. (1998). The role of inhibitory processes in young children’s
difficulties with deception and false belief. Child Development, 69, 672–691.
Cole, K., & Mitchell, P. (2000). Siblings in the development of executive control and a theory of
mind. British Journal of Developmental Psychology, 18, 279–295.
Cutting, A. L., & Dunn, J. (1999). Theory of mind, emotion understanding, language, and family
background: Individual differences and interrelations. Child Development, 70, 853–865.
Dunn, J. (1994). Changing minds and changing relationships. In C. Lewis & P. Mitchell (Eds.),
Origins of an understanding of mind ( pp. 297–310). Hove, UK: Erlbaum.
Dunn, J. (1995). Children as psychologists: The later correlates of individual differences in
understanding emotions and other minds. Cognition and Emotion, 9, 187–201.
Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test-Revised. Circle Pines, MN:
American Guidance Service.
Flavell, J. H. (1986). The development of children’s knowledge about the appearance-reality
distinction. American Psychologist, 41, 418–425.
Flavell, J. H. (1999). Cognitive development: Children’s knowledge about the mind. Annual
Review of Psychology, 50, 21–45.
Glass, D., Neulinger, J., & Brim, O. G. (1974). Birth order verbal intelligence and educational
aspirations. Child Development, 45, 807–811.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society
Mental playmates: Siblings, executive functioning and theory of mind
751
Gopnik, A., & Astington, J. W. (1988). Chidren’s understanding of representational change and its
relation to the understanding of false belief and the appearance-reality distinction. Child
Development, 59, 26–37.
Holmbeck, G. N. (1997). Toward terminological, conceptual and statistical clarity in the study of
mediators and moderators. Journal of Consulting and Clinical Psychology, 65, 599–610.
Hughes, C. (1998). Executive function in preschoolers: Links with theory of mind and verbal
ability. British Journal of Developmental Psychology, 16, 233–253.
Hughes, C., & Ensor, R. (2005). Theory of mind and executive function: A family affair?
Developmental Neuropsychology, 28, 645–668.
Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure associated with
theory of mind development in young children. Developmental Psychology, 32, 70–78.
Keppel, G. (1982). Design and analysis: A researcher’s handbook. Englewood Cliffs, NJ:
Prentice-Hall.
Lewis, C., Freeman, N. H., Kyriakidou, C., Maridaki-Kassotaki, K., & Berridge, D. M. (1996). Social
influences on false belief access: Specific sibling influences or general apprenticeship? Child
Development, 67, 2930–2947.
Perner, J., & Lang, B. (1999). Development of a theory of mind and cognitive control. Trends in
Cognitive Science, 3, 337–344.
Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You catch it from
your sibs. Child Development, 65, 1228–1238.
Peterson, C. C. (2000). Kindred spirits: Influence of siblings’ perspectives on theory of mind.
Cognitive Development, 15, 435–455.
Peterson, C. C., & Siegal, M. (1999). Representing inner worlds: Theory of mind in autistic, deaf
and normal-hearing children. Psychological Science, 10, 126–129.
Peterson, C. C., & Slaughter, V. P. (2003). Opening windows into the mind: Mothers’ preferences
for mental state explanations and children’s theory of mind. Cognitive Development, 18,
399–429.
Piaget, J. (1965). The moral judgment of the child. New York: Free Press.
Reed, M., Pein, D., & Rothbart, M. (1984). Inhibitory self-control in preschool children. MerrillPalmer Quarterly, 30, 131–147.
Repacholi, B., Slaughter, V., Pritchard, M., & Gibbs, V. (2003). Theory of mind, Machiavellianism
and social functioning in childhood. In B. Repacholi & V. Slaughter (Eds.), Individual
differences in theory of mind ( pp. 67–98). New York: Psychology Press.
Ruffman, T., Perner, J., Naito, M., Parkin, L., & Clements, W. A. (1998). Older (but not younger)
siblings facilitate false belief understanding. Developmental Psychology, 34, 161–174.
Slaughter, V. P., & Repacholi, B. (2003). Introduction: Individual differences in theory of mind:
What are we investigating? In B. Repacholi & V. Slaughter (Eds.), Individual differences in
theory of mind ( pp. 1–12). New York: Psychology Press.
Tabachnick, B., & Fidell, L. (1989). Using multivariate statistics. New York: Harper and Row.
Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind development:
The truth about false belief. Child Development, 72, 655–684.
Zajonc, R. B., Markus, H., & Markus, G. B. (1979). The birth order puzzle. Journal of Personality
and Social Psychology, 37, 1325–1341.
Received 1 June 2004; revised version received 1 August 2005