Content uploaded by Bruce D. Homer
Author content
All content in this area was uploaded by Bruce D. Homer on Jan 03, 2018
Content may be subject to copyright.
Developmental Trends in Flexibility and Automaticity of Social Cognition
Elizabeth O. Hayward
New York University
Bruce D. Homer
City University of New York
Manuel Sprung
Psychosomatisches Zentrum Waldviertel
Age-related changes in flexibility and automaticity of reasoning about social situations were investigated.
Children (N=101; age range =7;8–17;7) were presented with the flexibility and automaticity of social cogni-
tion (FASC), a new measure of social cognition in which cartoon vignettes of social situations are presented
and participants explain what is happening and why. Scenarios vary on whether the scenario is socially
ambiguous and whether or not language is used. Flexibility is determined by the number of unique, plausible
explanations, and automaticity is indicated by speed of response. Overall, both flexibility and automaticity
increased significantly with age. Language and social ambiguity influenced performance. Future work should
investigate differences in FASC in older populations and clinical groups.
Theory of mind (ToM) refers to one’s awareness of
the mental lives of others and is a central skill in
social cognition and social functioning more
broadly. In young children, mental-state reasoning
is related to social skill development (Watson,
Nixon, Wilson, & Capage, 1999), as well as whether
one is liked by others (Slaughter, Dennis, & Pritch-
ard, 2002). In middle childhood and adolescence,
ToM is involved in competently interacting with
peers (Bosacki & Astington, 1999), in controlling
how one is perceived (Banerjee & Yuill, 1999), and
in how one feels about the self (Bennett & Mat-
thews, 2000; Bosacki, 2000). Although some
researchers have argued for alternative accounts of
social understanding, including embodiment (e.g.,
Goldman & de Vignemont, 2009; Ratcliffe, 2007)
and direct perception (e.g., Gallagher, 2008),
research shows that cognitive processes such as
ToM play an important role in social functioning.
For example, recent longitudinal findings suggest
that advanced ToM understanding is related to
both prosocial behaviors and peer relations (Baner-
jee, Watling, & Caputi, 2011; Caputi, Lecce, Pagnin,
& Banerjee, 2012). In the cultures that have been
studied, first-order ToM reasoning is typically
achieved by age 5; however, the nature of mental-
state reasoning beyond the preschool years has not
been firmly established. It is generally accepted that
children are able to show competency in the sec-
ond-order false belief task (Perner & Wimmer,
1985), the most widely used measure of advanced
ToM, by 5 or 6 years of age (Miller, 2009). There is
cultural variation in the timing of various ToM
milestones, with greater variation for more
advanced understanding of the mind (see Slaughter
& Perez-Zapata, 2014).
Although most of the research has focused on
early developments, ToM continues to develop with
age (Miller, 2009). Evidence from cognitive neuro-
science indicates that the brain regions known to
play a critical role in mental-state reasoning con-
tinue to develop well into adolescence and adult-
hood (Dumontheil, Apperly, & Blakemore, 2010).
There is growing empirical support for the notion
that ToM development continues in middle child-
hood and adolescence. In particular, research has
established that age predicts ToM ability in children
and preadolescents, as assessed by faux pas recog-
nition, perspective taking, and social contextual
inference (Banerjee et al., 2011; Devine & Hughes,
2013; Dumontheil et al., 2010).
The authors would like to thank the individuals who partici-
pated in this research. We would also like to express our gratitude
to Katharine Pace, Seamus Donnelly, Yolanta Kornak, Megan
Bromley, Ruth Schwartz, and Kate Startz Barton for their assis-
tance with data collection, data entry, and manuscript proofing.
Correspondence concerning this article should be addressed to
Elizabeth O. Hayward, CREATE Lab, New York University, 196
Mercer Street, 8th floor, New York, NY, 10012. Electronic mail
may be sent to elizabeth.hayward@nyu.edu.
©2016 The Authors
Child Development ©2016 Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2016/xxxx-xxxx
DOI: 10.1111/cdev.12705
Child Development, xxxx 2016, Volume 00, Number 0, Pages 1–15
Currently available tasks conceptualize ToM in
many ways, though no single task explicitly cap-
tures the multiple factors at play in social cognitive
reasoning (Miller, 2012; Sprung, 2010). Many cur-
rent measures were originally developed for the
study of ToM in individuals with autism. One
broadly used measure, the strange stories (Happ!
e,
1994), includes various vignettes involving nonlit-
eral statements for which correct interpretation
requires inferring the speaker’s underlying motive
and intended meaning (e.g., joke, figure of speech,
irony, double bluff, etc.). There are several other
measures similar to the strange stories, including
the recently developed silent films task by Devine
and Hughes (2013)—afilm-based analog of the
strange stories task (for a review, see Miller, 2012).
One interpretation is that these tasks assess a
child’s ability to use familiar schemas (e.g., a girl
lying to her mother in saying the dog broke the
lamp) to facilitate inferences about behavior. These
schemas may be familiar to children from their
experiences or from narrative structures to which
children are exposed in books or other media. One
suggestion is that reasoning about social contexts,
such as the familiar stories used in the measures
described above, involve two different types of cog-
nitive processes: low-level processes that are cogni-
tively efficient but inflexible, and high-level
processes that are highly flexible but cognitively
demanding (Apperly, 2011; Homer, Halkitis, Moel-
ler, & Solomon, 2013). Accordingly, it is possible
that performance on tasks that involve high-level
processes becomes more routine with development.
However, stories such as those used in the strange
story paradigm and similar tasks seem to lend
themselves to one particular response or interpreta-
tion, such that the scoring schema rests on the
assumption of a correct interpretation. These vign-
ettes present situations that are familiar to children
in this culture and should therefore be relatively
unambiguous to interpret as children consider well-
rehearsed, “highly practiced”social events that may
or may not involve active cognitive processing
(Apperly & Butterfill, 2009). Although interpretation
of more familiar or unambiguous social vignettes
can rely on lower level, automatized processes, less
familiar, ambiguous vignettes require more active,
higher level social cognition. The comparison of
vignettes that presents a novel or ambiguous social
scenario to those that present a familiar or unam-
biguous scenario will be investigated in the current
study.
One task designed for use with typically devel-
oping preadolescents presents stories that are
purposefully ambiguous, such that they prompt a
range of interpretations (Bosacki & Astington,
1999). Interpreting such stories may involve more
complex processing than is used in processing more
familiar, unambiguous scenarios. Appropriately
interpreting ambiguous social scenarios is vital for
successful everyday social functioning (Klin, 2000).
Ambiguous stimuli have also been developed to
investigate ToM deficits among individuals with
high-functioning autism—namely, the Frith–Happ!
e
animations or triangles task (Abell, Happ!
e, & Frith,
2000). Although these animations are similar to
ambiguous social scenarios in that they are open-
ended, they are not suited for older groups, as there
is some evidence that, by at least 12 months of age,
typically developing individuals interpret the ani-
mations as recognizable social narratives (Abell
et al., 2000; Gergely, N!
adasdy, Csibra, & Biro, 1995;
Klin & Jones, 2006).
In the current study, we aim to investigate devel-
opmental trends in interpretations of both ambigu-
ous and more recognizable, familiar social scenarios
using a new tool, the FASC: flexibility and auto-
maticity of social cognition. The measure utilized in
the current study capitalizes on the conceptual
strengths of both the strange stories and the
ambiguous stories in presenting a range of social
scenarios in which beliefs and emotions drive
behavior. Given the receptive language demands in
both of these tasks in their original form, we devel-
oped both verbal and nonverbal vignettes in order
to investigate whether explicit language makes
these tasks easier or more difficult to interpret.
Therefore, the FASC consists of eight cartoons that
differ on two factors: ambiguity (ambiguous vs.
unambiguous social scenario) and language (pres-
ence vs. absence of language in the story; see Sup-
porting Information). We also introduced a novel
scoring schema in order to investigate developmen-
tal changes in automaticity and flexibility in reason-
ing about mental states (Apperly & Butterfill, 2009).
As children develop, it is plausible that they
become more efficient in their reasoning about
others’minds, as well as more flexible (Apperly &
Butterfill, 2009; Baron-Cohen, 1995). Apperly and
Butterfill (2009) argue that greater efficiency in
adult ToM reasoning may stem from automatiza-
tion: Adults’reasoning about the mental world
often involves highly familiar, well-practiced situa-
tions and habitual responses, or social scripts that
are followed with minimal cognitive effort (Bald-
win, 1992). This suggests that adults become effi-
cient as a result of learned associations, such that
expectations about future behavior are based on
2 Hayward, Homer, and Sprung
past experiences. On tasks of belief–desire reason-
ing and perspective taking, both response times and
errors decrease with age, indicating increased effi-
ciency and automaticity (Apperly, Warren, Andrews,
Grant, & Todd, 2011; Dumontheil et al., 2010). In an
effort to understand how this automaticity develops
over the course of childhood and adolescence, in the
current study we assess automaticity in ToM reason-
ing by measuring the time it takes to generate
hypotheses about mental-state information while
varying the presence of ambiguity and language in
the scenarios presented. On the basis of Apperly and
Butterfill’s theoretical framework, as well as previous
research (Hayward, 2011), we predict that adoles-
cents will be faster than children in their reasoning
about social situations, indicating an increase in
automaticity. We further expect that both familiarity
of narratives and the use of language in the narrative
will result in greater automaticity (i.e., faster process-
ing) of social scenarios.
It seems likely that as children grow into adoles-
cents and adults, they also become more competent
in generating multiple potential hypotheses to
explain others’behaviors, increasing in the flexibil-
ity in thinking about other’s mental states. Apperly
and Butterfill (2009) argue that once children mas-
ter the concept of belief, they become more flexible
in their ability to reason about mental states while
still lacking the efficiency of adults. Cognitively
flexible reasoning about social situations allows
adults to engage in what they term “top-down guid-
ance of social interaction,”or anticipation of others’
beliefs and intentions, as well as “explicit reasoning
about the causes and justifications for mental
thoughts”(p. 966). Therefore, this theory suggests
that children develop these more advanced social
cognitive skills with age and can apply them with
greater flexibility to novel or ambiguous situations.
The current study is designed to investigate whether
children become more flexible and more automatic
in reasoning about social situations between the ages
of 7 and 18 years, as Apperly and Butterfill’s theory
would suggest.
In the current study, flexibility of ToM reasoning
is assessed by measuring the number of plausible
hypotheses about mental states participants can
generate in response to social scenarios presented
as cartoons that differ in the factors of ambiguity
(ambiguous vs. unambiguous social scenario) and
language (presence vs. absence of language in the
story). In this way, responses are scored in a similar
manner as the Social Attribution Task, in which
individuals provide explanations for animations
depicting interactions between geometric shapes
(Klin, 2000; Klin & Jones, 2006). The Social Attribu-
tion Task was developed for use with clinical
groups, and typically developing individuals imme-
diately recognize the social nature of the task. Age
does not correlate with performance on this task,
suggesting that although this task is sensitive to
clinical variability, it is not a developmental mea-
sure (Klin & Jones, 2006). The current approach
allows for investigating whether there is develop-
ment of flexibility in children’s and adolescents’
social cognition, and whether this flexibility varies
with features of the scenario—in the current study,
the ambiguity of the scenarios and the presence or
absence of language. Familiar, unambiguous stories
by definition lend themselves to common or ready-
made interpretations; similarly, the use of language
may make certain specific interpretations of social
situations much more likely.
Although there are a number of measures that
capture developmental change in ToM during mid-
dle childhood (Banerjee et al., 2011; Devine &
Hughes, 2013; Dumontheil et al., 2010), the method-
ology presented in the current study (i.e., the
FASC) has several advantages over previously used
measures. The current methodology allows for
investigating developmental trends in both flexibil-
ity and automaticity in social reasoning, thereby
allowing us to assess the extent to which these
social cognitive skills develop as theory suggests.
Furthermore, the current methodology allows for a
comparison of children and adolescents’responses
to ambiguous versus familiar, less ambiguous situa-
tions, capitalizing on the strengths of previous mea-
sures. Unlike previous measures of advanced ToM,
the FASC also varies the receptive language
demands of the task by having some scenarios
with, and some without, language. Finally, the cur-
rent method goes beyond pass–fail data that over-
simplifies ToM competence (Repacholi & Slaughter,
2003). Current measures of advanced ToM assess
veridical ToM reasoning in which there exists a sin-
gular “true”answer, resulting in pass–fail data. In
contrast, the current measure assesses adaptive
ToM reasoning. Veridical decision making is
defined as reasoning with the assumption that there
is one intrinsically correct solution; in contrast,
adaptive decision making is reasoning with the
assumption that there is no intrinsically correct
solution but instead that there may be several plau-
sible solutions (Goldberg, 2009). Adaptive reasoning
is more akin to the type of reasoning individuals
do in most social interactions, where there cannot
be complete certainty about another person’s mind,
and beliefs about others’mental states may have to
Flexibility and Automaticity of Social Cognition 3
be modified as new information is acquired (Baron-
Cohen, 1995). In the current study, participants are
considered successful when they provide reasonable
hypotheses about the mental states of the characters
in the social scenarios. The current methodology
acknowledges the sticky truth about most social sit-
uations: We use modifiable hypotheses, rather than
fixed facts, to infer the mental states of others.
Language ability, in particular, receptive lan-
guage, has been associated with acquisition of ToM
(Apperly & Butterfill, 2009; Astington & Baird, 2005;
Lillard & Kavanaugh, 2014; Milligan, Astington, &
Dack, 2007). Research also suggests that language
and ToM continue to be related during middle child-
hood (Bosacki, 2000; Caputi et al., 2012) and beyond
(Newton & de Villiers, 2007); however, findings are
mixed (see Dungan & Saxe, 2012). Verbal abstract
reasoning ability has been found to predict perfor-
mance on the strange stories (Hayward, 2011) and
the silent films task (Devine & Hughes, 2013). It
seems clear that expressive language plays a role in
performance on tests such as the strange stories or
the ambiguous social scenarios. The verbal content in
the stories, however, likely influences individuals’
ability to reason about the mental states in those sto-
ries—the richer the verbal explanation provided in a
story, the more quickly and accurately participants
can make sense of the scenario. Therefore, it is neces-
sary that future tasks address the verbal content in
ToM stories specifically. In the current study,
although participants need to employ expressive lan-
guage in their responses, the task itself includes a
manipulation to assess the extent to which receptive
language plays a role in social cognitive reasoning.
In early childhood, there is some evidence for a
small gender difference in children’s false belief
understanding—with girls outperforming boys
(Charman, Ruffman, & Clements, 2002). Although
this effect is far from universal—for instance Hughes,
Ensor, and Marks (2011) did not find any significant
gender effects—previous research employing both
social vignettes, such as the faux pas test as well as
the ambiguous stories, with some consistency indi-
cate a female advantage in ToM reasoning in middle
childhood and adolescence (Baron-Cohen, O’Rior-
dan, Stone, Jones, & Plaisted, 1999; Bosacki & Asting-
ton, 1999; Devine & Hughes, 2013; Dumontheil et al.,
2010). Therefore, we also considered the effect of gen-
der on social cognitive reasoning in this age group.
The Current Study
The current cross-sectional study examines age-
related differences in social cognition in children
ages 7–18. To tease apart which aspects of social
cognitive reasoning develop over later childhood
and into adolescence, we investigated the develop-
ment of flexibility and automaticity in reasoning
about social situations while varying the level of
social ambiguity and use of language in the tasks.
The current methodology allowed for the compar-
ison of ambiguous to unambiguous or familiar
social scenarios, as well as verbal to nonverbal ver-
sions of tasks, in order to investigate within-indivi-
dual differences in processing social stimuli.
Specifically, we investigated the following research
questions:
Do Children and Adolescents Become More Flexible in
Their Ability to Generate Mental-State Explanations for
Social Behavior as They Age?
Hypothesis 1 is that children and adolescents
will become more flexible in providing explanations
for social behavior with age, as indicated by their
ability to generate multiple mental-state explana-
tions for both ambiguous and familiar unambigu-
ous social behavior (Apperly & Butterfill, 2009).
Do Children and Adolescents Become More Automatic
in Explaining Social Behavior as They Age?
Hypothesis 2 is that older children and adoles-
cents will be quicker than younger children in
generating mental-state explanations for both
ambiguous and familiar unambiguous social behav-
ior, indicating an age-related increase in automatic-
ity (Apperly & Butterfill, 2009; Dumontheil et al.,
2010).
How Does Ambiguity of a Social Scenario Affect
Flexibility and Automaticity in Generating Explanations
of Social Behavior?
Hypothesis 3 is that children and adolescents
will provide more mental-state explanations, using
more mental-state terms, in response ambiguous
scenarios, resulting in increased flexibility scores.
We also hypothesize that children will generate
mental-state explanations to familiar unambiguous
scenarios more quickly, and thus more automati-
cally, than to ambiguous scenarios (Apperly & But-
terfill, 2009). We predict that children and
adolescents will provide more conventional inter-
pretations in response to familiar unambiguous
schemas as compared to ambiguous scenarios and
that use of these conventional responses will
increase with age (Apperly & Butterfill, 2009).
4 Hayward, Homer, and Sprung
How Does the Use of Language Affect the Flexibility
and Automaticity in Generating Explanations of Social
Behavior?
Hypothesis 4 is that children and adolescents
will be more flexible and automatic in their
responses to verbal scenarios as compared to non-
verbal scenarios, generating more mental-state
explanations more quickly. Previous research has
found that language supports the acquisition of
ToM in children, and ToM in turn supports the
development of language (Apperly & Butterfill,
2009; Astington & Baird, 2005; Milligan et al., 2007).
Therefore, we hypothesize that children’s explana-
tions of social behaviors will be facilitated by lan-
guage in social scenarios, as language may stimulate
thinking about thoughts, beliefs, and desires of
others, resulting in increased flexibility scores such
as use of mental-state terms and mental-state
responses. However, alternatively, it is also plausible
that language may hinder flexibility in generating
diverse explanations, because language may make
the beliefs and desires of characters more explicit.
Furthermore, we predict that children will generate
explanations for verbal scenarios more quickly than
for nonverbal scenarios. However, it is also possible
that the presence of language in the verbal scenarios
may increase the amount of time it takes to generate
explanations, as it requires more active processing
(i.e., both the verbal and nonverbal information must
be processed and integrated to understand the sce-
nario). Language processing may require a greater
effort for the younger children, and therefore it again
seems likely that age may interact with the effects of
language in the social scenarios, resulting in greater
effects of language for younger children.
Method
Participants
Children (N=101), who ranged from 7;8 to
17;7 years of age, were recruited from a large urban
public school district in New York City. There were
52 (51.5%) female and 49 (48.5%) male in the sam-
ple. Participants were from a range of ethnic and
racial backgrounds: 2 identified as African Ameri-
can (2%), 5 as Asian (5%), 30 as Caucasian (29.7%),
and 44 as Latino (43.6%). For the remaining partici-
pants, 14 identified as “other,”and 6 participants
did not provide data on their ethnicity. Participants
were drawn from schools that varied from 15% to
96% in student eligibility to receive free lunch, sug-
gesting a range in socioeconomic status. Children
were required to have functional English language
skills to participate in the study. The percentage of
students who were English language learners at
participating schools ranged from 4% to 24%; all
participants identified as primary English speakers.
Given that participants were recruited from general
education classrooms, all students met the language
skills inclusion criteria. The data were collected dur-
ing the 2011–2012 school year. For some analyses,
participants were split into three groups: 8-year-
olds (7;8–9;10; n=29), 12-year-olds (11;4–14;4;
n=42), and 16-year-olds (16;5–17;7; n=30). These
age groups correspond roughly to preadolescence,
early adolescence, and middle adolescence, and
cover a period during which there is considerable
development in children’s social understanding
(Smetana, Campione-Barr, & Metzger, 2006). An a
priori power analysis, conducted using G*Power
(3.0.4; Faul, Erdfelder, Lang, & Buchner, 2007) soft-
ware, was run to indicate the necessary sample size
to lend sufficient power to the analyses. When con-
ducting multivariate analyses to identify large
effects (g
2
=.14), a target level of power of .80, and
an alpha level of a=.05, a sample of 58 partici-
pants is necessary, suggesting the current sample
offers sufficient power.
Measures
The primary stimuli consisted of eight vignettes
presented in cartoon format adapted from the Hap-
p!
e’s (1994) strange stories and Bosacki’s (2000)
socially ambiguous stories. The cartoons were cre-
ated with Bitstrips (www.bitstrips.com), using cus-
tomizable comic templates and characters. Bitstrips
were selected for use, as younger children easily
understand the combination of comics and simple
text, and the format is also familiar to adolescents,
as a means of expression often used in social media.
Cartoons depicted scenarios that varied in ambigu-
ity (ambiguous vs. unambiguous) and in the pres-
ence or absence of language accompanying the
cartoon (verbal vs. nonverbal). Therefore, two sto-
ries in each condition resulted in eight cartoons (see
Supporting Information for sample cartoons of each
story type). The unambiguous stories were based
on four vignettes featuring familiar schemas from
Happ!
e’s (1994) original strange stories, namely the
Misunderstanding (Bag), Lie (Dentist),Lie (Dog), and
White Lie (Hat). Participants saw all eight cartoons
in a random order. Participants were instructed,
“Here are some cartoons I would like you to look
at.”Before each image, they were also told, “Let
me know when you are finished.”When they
Flexibility and Automaticity of Social Cognition 5
indicated that they had finished looking at each pic-
ture and reading the simple text in the verbal car-
toons, the image was removed and the
experimenter asked, “Explain why [the character]
does what she or he does in this story.”After par-
ticipants offered their initial explanation aloud, they
were prompted to generate multiple explanations
for the character’s behavior using the prompt, “Can
you think of another reason?”This prompt was
repeated until the participant stated that they could
not think of any other reasons.
Responses were audio recorded and later tran-
scribed verbatim in InqScribe (Inquirium, LLC, Chi-
cago, IL, USA), a software for transcription. Use of
this software allowed for time stamping at the reso-
lution of one hundredth of a second. This procedure
was modeled on the scoring of previous open-ended
response tasks, such as Klin (2000). In order to quan-
tify the flexibility and automaticity of explanations
provided in response to each cartoon, we developed
a detailed coding system. First, the number of
responses to each cartoon was assessed. In most
cases, it was clearly evident when a participant was
providing a new response, as the new utterance was
stated in reaction to the prompt, “Can you think of
another reason?”However, if a participant sponta-
neously provided a new response, for example,
moving on to a new explanation by stating, “Or,”
before the examiner prompted for an additional
response, each response was scored separately. In
this way, the total number of responses was equal to
the total number of explanations provided by the
participant in response to the cartoon, whether
offered spontaneously or in response to the prompt.
In an effort to confirm that the unit of analysis of the
response was reliable, a second rater coded 25% of
the data. There was excellent agreement between
coders on the total number of responses recorded,
interclass correlation (ICC) =.996. Each response
was transcribed and coded as outlined below.
Four scores indicating flexibility and two scores
indicating automaticity were coded, resulting in a
total of six outcome scores.The flexibility scores
included:
Mental-State Terms
A comprehensive list of mental-state terms was
compiled based on previous work (Jenkins, Turrell,
Kogushi, Lollis, & Ross, 2003; LaBounty, Wellman,
Olson, Lagattuta, & Liu, 2008; Lagattuta & Well-
man, 2002). Given that the age range of the current
sample was broader than in previous work, mental-
state terms were added if they were used by one of
the participants but were not in the list of terms
that had been created from previous work (see Sup-
porting Information for the compiled mental-state
term list, with terms added by the authors noted in
boldface italics). We tallied the total number of
mental-state terms used, related to thought, desire,
or emotion, used across all responses. This score
was coded at the word level, whereas the remain-
ing flexibility scores were coded at the utterance
level. For example, consider the following response
to sample cartoon D (Supporting Information):
“Because the girl that’s alone playing with the ball
by herself looks sad and lonely and the two girls
felt bad.”This response would be given a score of
3 for the three mental-state terms used (sad, lonely,
felt bad). Self-referential mental-state terms used in
the context of answering (e.g., “I think,”“I don’t
know,”or “I guess”) were not coded as mental-
state terms. A second rater, blind to the age of the
participants, coded 25% of the data. There was
excellent agreement between coders on total men-
tal-state terms, ICC =.98.
Mental-State Response Score
Next, utterances were coded to assess the total
number of responses that involved unique mental
states, creating a sum of mental-state responses for
each cartoon. An utterance was coded as unique if
it included one or more of the following: A new
mental state (not just a synonym), a new motiva-
tion for the same mental state, or a new belief or
thought process attributed (even if mental-state
terms used were the same). For example, consider
the following response to sample cartoon D (Sup-
porting Information): “She doesn’t know how to
play so she—they went over there and I think they
tried to help her. Maybe the two girls, they’re
gonna tease her,”would receive a score of 2, as the
participant provided two motivations for the girls
pictured (helping and teasing). A second rater,
blind to the age of the participants, coded 25% of
the data. There was excellent agreement between
coders on total mental-state responses, ICC =.95.
Nonmental-State Response Score
This was defined as the tally of responses that
lacked mental-state reasoning. These utterances
tended to describe the pictures presented in cartoons
rather than infer motivations or emotions to the char-
acters. For example, in response to sample cartoon B,
one participant stated that, “Tom was a really good
player, but he had a weird body shape.”Responses
6 Hayward, Homer, and Sprung
that failed to provide any justification, such as stating
“I don’t know,”were also scored as nonmental-state
responses. This score was included to investigate
broader trends in age-related increases in responses,
as older children may use more physical and mental-
state responses overall. This score provided a point
of comparison to help identify developmental trends
that were unique to mental-state reasoning.
Conventional Responses
Responses to the FASC were also scored with
regard to the typicality or conventionality of the
interpretation provided in each utterance. Over the
course of transcribing, certain explanations for each
story came up repeatedly, across age groups. The
two most frequently occurring responses for each
story were flagged as conventional responses. For
example, in response to sample cartoon A, partici-
pants repeatedly stated, “He doesn’t want to hurt
Aunt Jane’s feelings/upset her”or “He doesn’t want
to be rude or wants to be polite.”In response to story
B, they often stated, “They wanted him to be on the
team or didn’t want him to be left out”or “They
want to pick on him/make fun of him.”Although
these responses can be considered the most likely or
plausible explanation for the cartoon, they were not
considered an indication of answering correctly ver-
sus incorrectly. Rather, they indicated the extent to
which the participant has answered similarly to his
or her peers. In the case of the ambiguous stories, in
which there is no clear “correct”interpretation for
what is occurring, conventional responses were con-
sidered similar to a “commonsense”logic rather than
“correct”reasoning. A set of two conventional
responses was generated per cartoon, based on the
high-frequency responses noted during transcription.
Therefore, participants received a score for whether
or not a conventional response was evident among
their responses for each story, receiving one score of
0 or 1 per story, for a total conventional response score
ranging from 0 to 2 for each story type and 0 to 8
across all eight stories. A second rater, blind to the
age of the participants, coded 25% of the data for
conventional responses. There was excellent agree-
ment between coders on total conventional response
scores, ICC =.92.
The automaticity scores for each participant
included the following.
Initial Reaction Time
This was defined as the time in seconds from the
beginning of the last word in the first prompt (e.g.,
for the prompt, “Explain why [the character] does
what she or he does in this story,”timing began at
“story;”Apperly et al., 2011) until the first word of
the participant’s initial utterance in making a gen-
uine response to the cartoon. (i.e., saying “um”or
“let me see,”etc., did not stop the timer). This score
gives a general sense of how much time a partici-
pant required to think of a response. Therefore,
there was an initial reaction time calculated for each
of the eight cartoons. Time was calculated using
timestamps marked in each InqScribe file.
Time Per Mental-State Response
This was defined as the time in seconds from
beginning of the last word in the first prompt to
the end of the last utterance divided by the number
of mental-state responses provided. This score gives
a sense of roughly how much time was required
for the participant to think of a mental-state
response in particular. For example, if a participant
provided two mental-state responses in 26 s, the
time per mental-state response would be 13 s.
Accordingly, this outcome variable is only available
for participants who provided mental-state
responses in the first place. Time was calculated
using timestamps marked in each InqScribe file.
Procedure
The study was conducted at a university
research laboratory. The protocols for the study
were approved by the university internal review
board and followed all APA guidelines for ethical
research. After obtaining parental consent and ver-
bal assent from the participants, participants were
tested individually for roughly 40 min in a quiet
room with an examiner. FASC cartoons were pre-
sented in 8 911 in. color printouts. Responses to
the FASC were audio recorded.
Results
After data screening and preliminary analyses, we
examined the effect of age and gender on multivari-
ate indicators of FASC. We then investigated uni-
variate effects of language, ambiguity, and age, as
well as interaction effects.
Data Screening
Each outcome variable was screened for outlying
scores falling >2SD above the overall mean;
Flexibility and Automaticity of Social Cognition 7
given the floor of 0 for this task, all outlying
scores fell above, rather than below, the mean. No
outliers were evident for the conventional response
score; the remaining outcome scores had between
four and eight outliers. Excluding outliers, the
sample used for each outcome was as follows:
mental-state terms, n=93; mental-state response
score, n=97; nonmental-state response score,
n=97; initial reaction time, n=95; and time per
mental-state response, n=94. The distribution of
each outcome variable was assessed at each age
level. Not surprisingly, the distribution for the fre-
quency variables of mental-state terms, mental-
state responses, and nonmental-state responses as
well as reaction time variables tended to be some-
what positively skewed. However, the skewness
and kurtosis were acceptable across all variables at
the three age levels with two exceptions. The
skewness and kurtosis statistics for the mental-
state responses for the 8-year-olds and the time
per mental-state response for 12-year-olds were
significantly greater than zero (p<.001), indicating
non normality. However, given that our sample
size provides no less than 20 cases in the smallest
cell, our analysis should be robust to violations of
multivariate normality (Mardia, 1971; Tabachnick
& Fidell, 2007).
Descriptive and Correlational Analyses
Descriptive statistics for the independent and
dependent variables involved in the analyses are
presented in Table 1. Correlations were corrected
for multiple comparisons using the Holm correc-
tion (Holm, 1979). Age as a continuous variable
was moderately positively associated with men-
tal-state responses, r(95) =.40, Holm-adjusted
p<.006 and the total number of conventional
responses, r(99) =.50, Holm-adjusted p<.006.
Age was negatively associated with the two
automaticity outcomes, initial reaction time, r
(93) =!.25, Holm-adjusted p=.026, and time per
mental-state response, r(92) =!.32, Holm-adjusted
p=.006. These preliminary analyses also revealed
moderate to strong associations among FASC
outcome scores, indicating a need for a multivari-
ate approach (Tabachnick & Fidell, 2007).
FASC Outcome Analyses
Flexibility of Social Cognition
Hypothesis 1 predicted that older participants
would demonstrate greater flexibility in their
responses to the FASC, as indicated by increased
scores on the FASC flexibility scores. Given the
associations among FASC flexibility scores, a multi-
variate analysis of variance (MANOVA) was con-
ducted to determine the effect of age on the
mental-state terms, mental-state responses, nonmen-
tal-state responses, and conventional responses
used. A one-way MANOVA with age group and
gender as between-subjects factors and mental-state
terms, mental-state responses, nonmental-state
responses, and conventional responses as depen-
dent measures, provided support for Hypothesis 1
—that flexibility increases with age—revealing a
main effect for age group, Wilks’k=.72, F(8,
164) =3.59, p=.001, g2
p¼:149. There was no mul-
tivariate main effect for gender, Wilks’k=.96, F(4,
82) =0.79, p=.538; therefore, gender was not
included in subsequent analyses. A series of three-
way repeated measures analyses of variance (ANO-
VAs) were conducted on each FASC score to fol-
low-up.
Mental-state terms. We first examined the use of
mental-state terms in responses to the cartoons.
Specifically, we examined Hypotheses 3 and 4,
whether ambiguous and verbal cartoons would
result in broader use of mental-state terms. Analy-
ses investigating participants’use of mental-state
terms indicated that, regardless of age, participants
produced more mental-state terms when describing
cartoons that depicted familiar, unambiguous social
scenarios but only when they were verbal as
opposed to nonverbal. Number of mental-state
terms was analyzed in a 3 (age group) 92 (ambi-
guity: ambiguous vs. unambiguous) 92 (language:
verbal vs. nonverbal) analysis of variance
(ANOVA), with age as a between-subjects variable
and ambiguity and language as within-subject vari-
ables. This analysis revealed main effects of both
ambiguity, F(1, 90) =32.47, mean square error
(MSE)=57.56, p<.001, g2
p¼:27, and language, F
(1, 90) =19.83, MSE =18.82, p<.001, g2
p¼:18.
These main effects were qualified by an interaction
between ambiguity and language, F(1, 90) =30.37,
MSE =39.55, p<.001, g2
p¼:25. Paired ttests indi-
cated that participants responded with significantly
more mental-state terms to unambiguous verbal
cartoons (M=3.22, SE =0.18, p<.001) than to all
other cartoon types. Unambiguous nonverbal car-
toons, M=2.10, SE =0.15, t(!2.60), p<.011, eli-
cited greater use of mental-state terms than did
ambiguous verbal cartoons (M=1.75, SE =0.13)
but were equivalent to ambiguous nonverbal car-
toons, M=1.95, SE =0.14, t(!1.72), p=.09. There
was no difference in the number of mental-state
8 Hayward, Homer, and Sprung
terms used in ambiguous nonverbal as compared to
unambiguous nonverbal cartoons. There was no
main effect for age on mental-state terms, F(2,
87) =0.47, MSE =1.90, p=.627. Therefore,
although participants’use of mental-state terms did
not increase overall with age, children and adoles-
cents used more mental-state terms in response to
unambiguous verbal scenarios than other scenarios.
Mental-state responses. Next we investigated
trends in mental-state responses to the cartoons.
Specifically, we examined Hypothesis 3, whether
ambiguous cartoons would result in a greater num-
ber of mental-state responses, and Hypothesis 4,
that the use of language in cartoons would result in
more mental-state responses. Analyses of the num-
ber of mental-state responses again suggest that
verbal unambiguous cartoons prompted a greater
number of mental-state interpretations. A three-way
repeated measures 3 (age group) 92 (ambiguity:
ambiguous vs. unambiguous) 92 (language: verbal
vs. nonverbal) ANOVA was conducted with total
mental-state responses as the dependent variable.
There were main effects for both ambiguity, F(1,
94) =19.29, MSE =7.83, p<.001, g2
p¼:17, and lan-
guage, F(1, 94) =9.83, MSE =2.72, p=.002,
g2
p¼:10. Again, these main effects were qualified
by a significant interaction between ambiguity and
language, F(1, 94) =7.95, MSE =1.81, p=.006,
g2
p¼:08. Paired ttests indicated that participants
provided more mental-state explanations to unam-
biguous verbal cartoons (M=1.61, SE =0.08,
p<.001) relative to all other cartoon types. The
number of mental-state responses provided for
ambiguous verbal cartoons (M=1.19, SE =0.07),
unambiguous nonverbal cartoons (M=1.30,
SE =0.07), and ambiguous nonverbal cartoons
(M=1.16, SE =0.07) were statistically equivalent.
Thus, these findings did not substantiate Hypothe-
sis 3 but did provide support for Hypothesis 4.
Age impacted the number of mental-state
responses provided, lending further support to
Hypothesis 1. As can be seen in Figure 1, the num-
ber of mental-state interpretations increased with
age. The analysis revealed a main effect for age
group, F(2, 94) =7.73, MSE =8.65, p=.001,
g2
p¼:14. Least significant difference test (LSD) post
hoc comparisons indicate that the number of men-
tal-state responses increased incrementally, with 8-
year-olds providing the fewest (M=1.06,
SE =0.10), 12-year-old providing somewhat more
(M=1.31, SE =0.08; p=.057), and 16-year-olds
providing significantly more mental-state responses
than both younger groups (M=1.62, SE =0.10;
p<.001).
Table 1
Means, Standard Deviation, and Bivariate Correlations Between Measures
8 years
M(SD)
12 years
M(SD)
16 years
M(SD)
Mental-state terms
(N=92)
Mental-state
response
(N=97)
Nonmental-state
response
(N=98)
Conventional
responses
(N=101)
Initial RT
(N=95)
Mental-state
response time
(N=94)
Age (continuous variable) 8.49 (0.60) 12.63 (0.95) 16.99 (0.31) .12 .40** .18 .50** !.25* !.32*
Total mental-state terms 2.12 (0.86) 2.26 (1.12) 2.39 (1.00) —.79** .27* .30* !.11 !.28*
Total mental-state responses 1.06 (0.37) 1.28 (0.49) 1.49 (0.61) —.43** .47** !.12 .13
Total nonmental-state responses 0.42 (0.24) 0.47 (.28) 0.52 (0.34) —.15 .05 .24*
Total conventional responses 3.59 (1.59) 4.38 (1.81) 6.00 (1.44) —!.19 !.13
Average initial reaction time (s) 2.24 (0.90) 1.76 (0.76) 1.65 (0.69) —.30*
Average mental-state
response time (s)
19.05 (6.97) 12.66 (6.15) 12.52 (5.85) —
*Holm-adjusted p<.05. **Holm-adjusted p≤.01.
Flexibility and Automaticity of Social Cognition 9
Nonmental-state responses. When examining use
of nonmental-state utterances, that simply described
the pictorial information in the cartoon or failed to
give any justification in response, the results suggest
that participants tended to provide the fewest non-
mental-state responses in reaction to verbal cartoons
that used familiar schemas. This is of interest as it
suggests that mental-state responses may have been
more readily accessed when a familiar scenario was
presented. A three-way repeated measures 3 (age
group) 92 (ambiguity: ambiguous vs. unambigu-
ous) 92 (language: verbal vs. nonverbal) ANOVA
was conducted with nonmental-state responses as
the dependent variable. There were main within-
subjects effects for both ambiguity, F(1, 91) =37.24,
MSE =7.55, p<.001, g2
p¼:29, and language, F(1,
91) =8.53, MSE =1.75, p=.004, g2
p¼:09. These
main effects were once again qualified by a signifi-
cant interaction between ambiguity and language, F
(1, 91) =17.57, MSE =2.89, p<.001, g2
p¼:16.
Paired ttests indicated that participants provided
fewer nonmental-state explanations to unambiguous
verbal cartoons (M=0.17, SE =0.03, p<.001) rela-
tive to all other cartoon types. The number of non-
mental-state responses provided for ambiguous
verbal cartoons (M=0.63, SE =0.06), unambiguous
nonverbal cartoons (M=0.48, SE =0.05), and
ambiguous nonverbal cartoons (M=0.59,
SE =0.05) were statistically equivalent.
With regard to the between-groups factor of age
group, no main effect was evident. The develop-
mental trend identified in mental-state responses
does not hold true for nonmental-state responses.
Given that the nonmental-state responses have the
same expressive language demands as mental-state
responses, these findings suggest that the observed
age-related improvements may be due to develop-
ment of mental-state reasoning rather than develop-
ments in general language or reasoning ability.
Conventional responses. With regard to conven-
tional responses, all participants provided more
conventional responses or common interpretations
to unambiguous cartoons that were accompanied
by language; the presence of language was espe-
cially influential for the youngest participants. This
was evident in the results from a three-way
repeated measures 3 (age group) 92 (ambiguity:
ambiguous vs. unambiguous) 92 (language: verbal
vs. nonverbal) ANOVA conducted with conven-
tional responses as the dependent variable. This
analysis revealed within-subjects main effects for
ambiguity, F(1, 95) =56.53, MSE =22.24, p<.001,
g2
p¼:37, and language, F(1, 95) =46.40,
MSE =16.80, p<.001, g2
p¼:33. Overall, partici-
pants provided more conventional responses for
unambiguous cartoons (M=1.40, SE =0.05) than
ambiguous cartoons (M=0.92, SE =0.05), provid-
ing support for Hypothesis 3 that familiar schemas
Figure 1. Mental-state response scores and conventional response scores in the flexibility and automaticity of social cognition by age
group.
10 Hayward, Homer, and Sprung
result in conventional interpretations. Participants
also provided more conventional responses for ver-
bal cartoons (M=1.37, SE =0.05) than nonverbal
(M=0.95, SE =0.05).
There was also a significant Language 9Age
Group interaction, F(2, 95) =3.67, MSE =1.33,
p=.029, g2
p¼:07. When the groups were analyzed
separately, the effect of language on conventional
response scores was significant for all age groups,
though stronger for 8-year-olds, F(1, 27) =32.30,
MSE =12.51, p<.001, g2
p¼:55 (M
verbal
=1.22,
SE
verbal
=0.10; M
nonverbal
=0.57, SE
nonverbal
=0.08),
as compared to 12-year-olds, F(1, 40) =10.08,
MSE =4.55, p=.003, g2
p¼:20 (M
verbal
=1.27, SE
ver-
bal =0.08; Mnonverbal
=0.94, SE
nonverbal
=0.07), and
16-year-olds, F(1, 28) =9.58, MSE =2.01, p=.004,
g2
p¼:25 (M
verbal
=1.61, SE
verbal
=0.10; M
nonver-
bal =1.35, SEnonverbal
=0.08). In sum, the effect of lan-
guage in facilitating common or conventional
responses was most pronounced for the youngest
participants.
Between-subjects analyses indicated that older
participants provided more typical or conventional
responses than younger participants, lending sup-
port for Hypothesis 1, as there was a main effect
for age, F(2, 98) =16.74, MSE =11.31, p<.001,
g2
p¼:26. LSD post hoc comparisons again indicate
that the number of conventional responses
increased with each age group, with 8-year-olds
providing the fewest (M=0.90, SE =0.08), 12-year-
olds providing significantly more (M=1.10,
SE =0.06; p=.048), and 16-year-olds providing the
greatest number of conventional responses
(M=1.5, SE =0.07; p<.001; see Figure 1.)
Automaticity of Social Cognition
As the FASC automaticity outcome scores were
moderately correlated with one another, a MAN-
OVA was conducted to determine the multivariate
effect of age on the initial reaction time and time
per mental-state response. A one-way MANOVA
with age group and gender as between-subjects fac-
tors and initial reaction time and time per mental-
state response scores as dependent measures indi-
cated that automaticity increases with age, as pre-
dicted in Hypothesis 2. The analysis revealed a
significant main effect for age, Wilks’k=.80, F(4,
170) =4.97, p=.001, g2
p¼:10. There was no multi-
variate main effect for gender, Wilks’k=.98, F(2,
85) =.75, p=.475; therefore, gender was not
included in subsequent analyses. A series of three-
way repeated measures ANOVAs were conducted
on each FASC score to follow up.
Initial reaction time. Analyses examining how
long it took participants to generate an initial inter-
pretation indicated that the within-group factor of
ambiguity was a significant factor, and there was
also evidence of an effect of age. A three-way
repeated measures 3 (age group) 92 (ambiguity:
ambiguous vs. unambiguous) 92 (language: verbal
vs. nonverbal) ANOVA conducted with initial reac-
tion time as the dependent variable indicated a
within-subjects main effect for ambiguity, F(1,
92) =7.22, MSE =11.41, p=.009, g2
p¼:07. Partici-
pants took longer to respond to ambiguous cartoons
(M=2.06, SE =0.11) than unambiguous cartoons
(M=1.71, SE =0.10). Therefore, Hypothesis 3 that
ambiguity would result in slower responses was con-
firmed, though Hypothesis 4, that language would
result in quicker responses, was not supported.
Hypothesis 2, that automaticity of reasoning
increases with age, was supported, as the effect for
age group on the initial response time was also sig-
nificant, F(2, 92) =4.55, MSE =11.04, p=.013,
g2
p¼:09. LSD post hoc comparisons indicated that
the 8-year-olds took longer (M=2.24, SE =0.15)
relative to the 12-year-olds (M=1.76, SE =0.12;
p=.017) and the 16-year-olds (M=1.65, SE =0.14;
p=.005). The difference in the initial response time
between 12- and 16-year-olds was marginally sig-
nificant (p=.054).
Time per mental-state response. Similarly, ambigu-
ity negatively impacted the time it took participants
to generate mental-state explanations in response to
the cartoons, but language had no effect. A total of
79 participants (twenty-three 8-year-olds, twenty-
eight 12-year-olds, and twenty-eight 16-year-olds)
gave mental-state responses to all story types, pro-
viding a basis for investigating within-subjects dif-
ferences in the time per mental-state response score.
A three-way repeated measures 3 (age group) 92
(ambiguity: ambiguous vs. unambiguous) 92 (lan-
guage: verbal vs. nonverbal) ANOVA conducted
with the time per mental-state response as the
dependent variable. With regard to within-subjects
effects, there was a main effect for ambiguity, F(1,
70) =6.98, MSE =124.30, p=.010, g2
p¼:09. Partici-
pants required more time to generate mental-state
responses to ambiguous cartoons (M=15.57,
SE =0.71) than unambiguous cartoons (M=14.24,
SE =0.76). Therefore, ambiguity negatively
impacted the time taken to produce mental-state
responses as predicted in Hypothesis 3, though
again, Hypothesis 4, that language would increase
automaticity, was not supported.
These analyses also suggest that age played a
role in the time it took participants to generate
Flexibility and Automaticity of Social Cognition 11
mental-state responses, supporting Hypothesis 2.
Results indicated a main effect for age group, F(2,
70) =11.30, MSE =1,521.70, p<.001, g2
p¼:24,
such that 8-year-olds (M=19.87, SE =1.37) took
longer to respond with a mental-state explanation
than both the 12- and 16-year-olds (M=11.96,
SE =1.12 and M=12.89, SE =1.10, respectively,
p<.001). The 12- and 16-year-olds did not differ in
their response time when generating mental-state
explanations (p=.555).
Discussion
The purpose of this study was to investigate devel-
opments in flexibility and automaticity in reasoning
about the social world in late childhood and adoles-
cence. Overall, our findings suggest that children
and adolescents become more flexible and efficient
in their reasoning about mental states, as most of
the outcome scores from the FASC improved with
age. In terms of flexibility, there was a significant
effect of age in the multivariate analyses, providing
support for Hypothesis 1 that flexibility increases
with age. Further analyses revealed that the adoles-
cents were able to generate significantly more men-
tal-state responses and conventional responses as
compared to younger children, but there were no
age-related differences in the number of nonmental-
state responses. This casts doubt on the interpreta-
tion that the developmental effects we observed can
be attributed to expressive language in general
rather than mental-state reasoning in particular.
With regard to automaticity, the analyses supported
Hypothesis 2 that children and adolescents show
greater automaticity in mental-state reasoning with
age. Further analyses revealed that with age, chil-
dren grow faster in reasoning about mental states,
as indicated by their initial response time and time
per mental-state response in the FASC. Therefore,
consistent with recent theoretical work (Apperly &
Butterfill, 2009), our findings suggest that as chil-
dren age through early adolescence, they become
more capable in providing explanations for social
behavior, increasing in both flexibility and auto-
maticity.
Ambiguity and Language in Social Cognition
Furthermore, we were interested in investigating
the effect of the use of familiar unambiguous sche-
mas and language in facilitating social cognition
across later childhood and adolescence. Contrary to
our expectation, children provided a greater
number of mental-state explanations when faced
with unambiguous scenarios as compared to
ambiguous scenarios. However, there was an effect
of ambiguity on automaticity, as it took children
more time to generate responses to ambiguous sce-
narios, suggesting greater cognitive effort. Regard-
ing flexibility, children and adolescents generated a
greater number of conventional responses in the
unambiguous, familiar scenarios. These findings
support Hypothesis 3, that the degree of ambiguity
in a social scenario affects both the flexibility and
efficiency of mental-state reasoning. The findings
suggest that children and adolescents are better
able to reason about social scenarios that present
familiar themes, as they generated more mental-
state explanations in response to unambiguous
scenarios. More conventional interpretations were
provided when children and adolescents were faced
with unambiguous, familiar scenarios as compared
to ambiguous cartoons, suggesting that these sce-
narios lend themselves to inflexible, routine, or
“stock,”responses across age groups, as suggested
by Apperly (2011).
When language was used in the scenarios pre-
sented, participants provided a greater number of
mental-state explanations, providing partial support
for Hypothesis 4. This was particularly true for the
verbal unambiguous scenarios. Additionally, chil-
dren and adolescents were able to generate more
mental-state terms and give more mental-state
responses for the verbal unambiguous scenarios
compared to the nonverbal ambiguous scenarios.
Participants also generated significantly more con-
ventional responses for the verbal scenarios, sug-
gesting that the use of language may prompt
individuals to use established scripts to interpret a
social situation. These findings are consistent with
theories that suggest language and familiar schemas
provide in-roads to ToM development (Apperly &
Butterfill, 2009; Astington & Baird, 2005; Bosacki,
2000).
Limitations
With regard to limitations, first, although the
current results suggest age-related change in social
cognitive reasoning, as older children offered a
greater number of mental-state responses compared
to younger children, we cannot rule out the possi-
bility that these results are due to age-related
increases in verbal fluency that are not unique to
social cognitive development. Future work with
these stimuli should employ cartoons depicting a
nonsocial scenario as a control to rule out this more
12 Hayward, Homer, and Sprung
general explanation of our findings. Second,
although we sought to control for language
demands within the task, there were nonetheless
expressive language demands in the response for-
mat. Therefore, though the role of language was
diminished by design, the nonverbal stories cannot
be considered a language-free format. Nonverbal
versions of ToM tasks necessitate the use of a picto-
rial multiple-choice response format, with a clear
correct versus incorrect response (Krachun, Carpen-
ter, Call, & Tomasello, 2010); given the current goal
of employing a measure that captures adaptive
advanced ToM reasoning, this response format was
not possible. Future research should include a mea-
sure of expressive language to help account for lan-
guage effects. Third, given that we purposefully
included familiar narrative schemas within the
materials, these findings may very well be limited
to the culture from which they originate. Our
familiar, unambiguous scenarios are likely to feel
far more ambiguous and less readily interpreted to
individuals from disparate cultures. This is of con-
siderable interest and warrants cross-cultural fol-
low-up research. The generalizability of findings
may also be limited by environmental factors that
influence ToM such as family structure. Fourth,
the current study investigated a limited age range,
spanning middle childhood and adolescence with
a relatively small sample size. Future work should
include larger sample sizes and broader age range
so as to make clear the developmental trajectory of
advanced social cognitive skills across the life
span.
Conclusions and Future Directions
Results provide theoretical and empirical support
for developmental change in FASC over middle
childhood and adolescence. Future work will focus
on investigating variability in flexibility and auto-
maticity with broader age groups and across cul-
tures, as well as within clinical populations.
Advanced ToM research with traditional tasks in
atypical adult populations has yielded inconsistent
findings (Br€
une, 2005; Happ!
e, 1994; Harrington, Sie-
gert, & McClure, 2005; Homer et al., 2013). Future
research investigating flexibility and automaticity in
social cognition may help illuminate subtle individ-
ual differences in social cognitive reasoning across
adult clinical populations. Among typically devel-
oping children, future work should also focus on
the educational implications of these changes in
flexibility and automaticity, as it seems plausible
that gains in these areas may impact social behavior
in the school environment, which may in turn
impact academic performance.
In conclusion, the current study provides further
insight into later developments in children and ado-
lescents’ToM. It suggests that two factors in social
cognition, flexibility and automaticity, continue to
develop in adolescence.
References
Abell, F., Happ!
e, F., & Frith, U. (2000). Do triangles play
tricks? Attribution of mental states to animated shapes
in normal and abnormal development. Cognitive Devel-
opment,15,1–16. doi:10.1016/S0885-2014(00)00014-9
Apperly, I. (2011). Mindreaders: The cognitive basis of theory
of mind. Hove, UK and New York, NY: Psychology
Press.
Apperly, I. A., & Butterfill, S. A. (2009). Do humans have
two systems to track beliefs and belief-like states? Psy-
chological Review,116, 953–970. doi:10.1037/a0016923
Apperly, I. A., Warren, F., Andrews, B. J., Grant, J., &
Todd, S. (2011). Developmental continuity in theory of
mind: Speed and accuracy of belief-desire reasoning in
children and adults. Child Development,82, 1691–1703.
doi:10.1111/j.1467-8624.2011.01635.x
Astington, J. W., & Baird, J. (Eds.). (2005). Why language
matters for theory of mind. New York, NY: Oxford
University Press.
Baldwin, M. W. (1992). Relational schemas and the pro-
cessing of social information. Psychological Bulletin,112,
461. doi:10.1037/0033-2909.112.3.461
Banerjee, R., Watling, D., & Caputi, M. (2011). Peer rela-
tions and the understanding of faux pas: Longitudinal
evidence for bidirectional associations. Child Development,
82, 1887–1905. doi:10.1111/j.1467-8624.2011.01669.x
Banerjee, R., & Yuill, N. (1999). Children’s understanding
of self-presentational display rules: Associations with
mental state understanding. British Journal of Develop-
mental Psychology,17, 111–124. doi:10.1348/
026151099165186
Baron-Cohen, S. (1995). Mindblindness: An essay on autism
and theory of mind. Cambridge, MA: The MIT Press.
Baron-Cohen, S., O’Riordan, M., Stone, V., Jones, R., &
Plaisted, K. (1999). Recognition of faux pas by normally
developing children with Asperger syndrome or high-
functioning autism. Journal of Autism and Developmental
Disorders,29, 407–418. doi:10.1023/A:1023035012436
Bennett, M., & Matthews, L. (2000). The role of second-
order belief-understanding and social context in chil-
dren’s self-attribution of social emotions. Social Develop-
ment,9, 126–130. doi:10.1111/1467-9507.t01-1-00115
Bosacki, S. L. (2000). Theory of mind and self-concept in
preadolescents: Links with gender and language. Jour-
nal of Educational Psychology,92, 709–717. doi:10.1037/
0022-0663.92.4.709
Bosacki, S., & Astington, J. W. (1999). Theory of mind in
preadolescence: Relations between social understanding
Flexibility and Automaticity of Social Cognition 13
and social competence. Social Development,8, 237–255.
doi:10.1111/1467-9507.00093
Br€
une, M. (2005). “Theory of mind”in schizophrenia: A
review of the literature. Schizophrenia Bulletin,31, 21–42.
doi:10.1093/schbul/sbi002
Caputi, M., Lecce, S., Pagnin, A., & Banerjee, R. (2012).
Longitudinal effects of theory of mind on later peer
relations: The role of prosocial behavior. Developmental
Psychology,48, 257–270. doi:10.1037/a0025402
Charman, T., Ruffman, T., & Clements, W. A. (2002). Is
there a gender difference in false belief development?
Social Development,11,1–10. doi:10.1111/1467-9507.
00183
Devine, R. T., & Hughes, C. (2013). Silent films and
strange stories: Theory of mind, gender, and social
experiences in middle childhood. Child Development,84,
989–1003. doi:10.1111/cdev.12017
Dumontheil, I., Apperly, I. A., & Blakemore, S. (2010).
Online usage of theory of mind continues to develop in
late adolescence. Developmental Science,13, 331–338.
doi:10.1111/j.1467-7687.2009.00888.x
Dungan, J., & Saxe, R. (2012). Matched false belief perfor-
mance during verbal and nonverbal interference. Cogni-
tive Science,36, 1148–1156. doi:10.1111/j.1551-6709.2012.
01248.x
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007).
G*Power 3: A flexible statistical power analysis pro-
gram for the social, behavioral, and biomedical
sciences. Behavior Research Methods,39, 175–191. doi:10.
3758/BF03193146
Gallagher, S. (2008). Direct perception in the intersubjec-
tive context. Consciousness and Cognition,17, 535–543.
doi:10.1016/j.concog.2008.03.003
Gergely, G., N!
adasdy, Z., Csibra, G., & Biro, S. (1995).
Taking the intentional stance at 12 months of age. Cog-
nition,56, 165–193. doi:10.1016/0010-0277(95)00661-H
Goldberg, E. (2009). The new executive brain: Frontal lobes
in a complex world. New York, NY: Oxford University
Press.
Goldman, A., & de Vignemont, F. (2009). Is social cogni-
tion embodied? Trends in Cognitive Sciences,13, 154–
159. doi:10.1016/j.tics.2009.01.007
Happ!
e, F. G. (1994). An advanced test of theory of mind:
Understanding of story characters’thoughts and feelings
by able autistic, mentally handicapped, and normal chil-
dren and adults. Journal of Autism and Developmental
Disorders,24, 129-154. doi:10.1007/BF02172093
Harrington, L., Siegert, R. J., & McClure, J. (2005). Theory
of mind in schizophrenia: A critical review. Cognitive
Neuropsychiatry,10, 249–286. doi:10.1080/13546800444
000056
Hayward, E. O. (2011). Measurement of advanced theory of
mind in school-age children: Investigating the validity of a
unified construct. Unpublished Doctoral Dissertation,
New York University, New York, NY.
Holm, S. (1979). A simple sequentially rejective multiple
test procedure. Scandinavian Journal of Statistics,6, 65–
70. http://www.jstor.org/stable/4615733
Homer, B. D., Halkitis, P. N., Moeller, R. W., & Solomon,
T. M. (2013). Methamphetamine use and HIV in rela-
tion to social cognition. Journal of Health Psychology,7,
900–910. doi:10.1177/1359105312457802
Hughes, C., Ensor, R., & Marks, A. (2011). Individual dif-
ferences in false belief understanding are stable from 3
to 6 years and predict children’s mental state talk with
school friends. Journal of Experimental Child Psychology,
108, 96–112. doi:10.1016/j.jecp.2010.07.012
Jenkins, J. M., Turrell, S. L., Kogushi, Y., Lollis, S., &
Ross, H. S. (2003). A longitudinal investigation of the
dynamics of mental state talk in families. Child Develop-
ment,74, 905–920. doi:10.1111/1467-8624.00575
Klin, A. (2000). Attributing social meaning to ambiguous
visual stimuli in higher-functioning autism and Asper-
ger syndrome: The social attribution task. Journal of
Child Psychology and Psychiatry,41, 831–846. doi:10.
1111/1469-7610.00671
Klin, A., & Jones, W. (2006). Attributing social and physi-
cal meaning to ambiguous visual displays in individu-
als with higher-functioning autism spectrum disorders.
Brain and Cognition,61, 40–53. doi:10.1016/j.bandc.2005.
12.016
Krachun, C., Carpenter, M., Call, J., & Tomasello, M.
(2010). A new change-of-contents false belief test: Chil-
dren and chimpanzees compared. International Journal
of Comparative Psychology,23, 145–165.
LaBounty, J., Wellman, H. M., Olson, S., Lagattuta, K., &
Liu, D. (2008). Mothers’and fathers’use of internal
state talk with their young children. Social Development,
17, 757–775. doi:10.1111/j.1467-9507.2007.00450.x
Lagattuta, K. H., & Wellman, H. M. (2002). Differences in
early parent-child conversations about negative versus
positive emotions: Implications for development of psy-
chological understanding. Developmental Psychology,38,
564–580. doi:10.1037//0012-1649.38.4.564
Lillard, A. S., & Kavanaugh, R. D. (2014). The contribu-
tion of symbolic skills to the development of an explicit
theory of mind. Child Development,85, 1535–1551.
doi:10.1111/cdev.12227
Mardia, K. V. (1971). The effect of nonnormality on some
multivariate tests and robustness to nonnormality in
the linear model. Biometrika,58, 105–121. doi:10.1093/
biomet/58.1.105
Miller, S. A. (2009). Children’s understanding of second-
order mental states. Psychological Bulletin,135, 749–773.
doi:10.1037/a0016854
Miller, S. A. (2012). Theory of mind: Beyond the preschool
years. New York, NY: Psychology Press.
Milligan, K., Astington, J. W., & Dack, L. A. (2007). Lan-
guage and theory of mind: Meta-analysis of the relation
between language ability and false belief understand-
ing. Child Development,78, 622–646. doi:10.1111/j.1467-
8624.2007.01018.x
Newton, A. M., & de Villiers, J. G. (2007). Thinking while
talking: Adults fail nonverbal false belief reasoning.
Psychological Science,18, 574–579. doi:10.1111/j.1467-
9280.2007.01942.x
14 Hayward, Homer, and Sprung
Perner, J., & Wimmer, H. (1985). “John thinks that Mary
thinks that . . .”: Attribution of second-order beliefs by
5- to 10-year-old children. Journal of Experimental Child
Psychology,39, 437–471. doi:10.1016/0022-0965(85)
90051-7
Ratcliffe, M. (2007). Rethinking commonsense psychology: A
critique of folk psychology, theory of mind and simulation.
New York, NY: Palgrave Macmillan.
Repacholi, B., & Slaughter, V. (Eds.). (2003). Individual dif-
ferences in theory of mind: Implications for typical and atyp-
ical development. New York, NY: Psychology Press.
Slaughter, V., Dennis, M. J., & Pritchard, M. (2002). The-
ory of mind and peer acceptance in preschool children.
British Journal of Developmental Psychology,20, 545–564.
doi:10.1348/026151002760390945
Slaughter, V., & Perez-Zapata, D. (2014). Cultural varia-
tions in the development of mind reading. Child Devel-
opment Perspectives,8, 237–241. doi:10.1111/cdep.1209
Smetana, J. G., Campione-Barr, N., & Metzger, A. (2006).
Adolescent development in interpersonal and societal
contexts. Annual Review of Psychology,57, 255–284.
doi:10.1146/annurev.psych.57.102904.190124
Sprung, M. (2010). Clinically relevant measures of chil-
dren’s theory of mind and knowledge about thinking:
Non-standard and advanced measures. Child and Ado-
lescent Mental Health,15, 204–216. doi:10.1111/j.1475-
3588.2010.00568.x
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate
statistics. Boston, MA: Pearson.
Watson, A. C., Nixon, C. L., Wilson, A., & Capage, L.
(1999). Social interaction skills and theory of mind in
young children. Developmental Psychology,35, 386–391.
doi:10.1037/0012-1649.35.2.386
Supporting Information
Additional supporting information may be found in
the online version of this article at the publisher’s
website:
Data S1. Flexibility and Automaticity of Social
Cognition (FASC) Sample Cartoons
Data S2. Mental-State Terms
Flexibility and Automaticity of Social Cognition 15