671845

research-article2016
AUT0010.1177/1362361316671845AutismDean et al.

Special Issue Article

Autism

The art of camouflage: Gender differences 2017, Vol. 21(6) 678­–689

© The Author(s) 2016
Reprints and permissions:
in the social behaviors of girls and boys sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1362361316671845
https://doi.org/10.1177/1362361316671845

with autism spectrum disorder journals.sagepub.com/home/aut

Michelle Dean1, Robin Harwood2 and Connie Kasari3

Abstract
This study examined the extent to which gender-related social behaviors help girls with autism spectrum disorder to
seemingly mask their symptoms. Using concurrent mixed methods, we examined the social behaviors of 96 elementary
school children during recess (autism spectrum disorder = 24 girls and 24 boys, typically developing = 24 girls and 24
boys). Children with autism spectrum disorder had average intelligence (IQ ⩾ 70), a confirmed diagnosis, and were
educated in the general education classroom. Typically developing children were matched by sex, age, and city of
residence to children with autism spectrum disorder. The results indicate that the female social landscape supports
the camouflage hypothesis; girls with autism spectrum disorder used compensatory behaviors, such as staying in close
proximately to peers and weaving in and out of activities, which appeared to mask their social challenges. Comparatively,
the male landscape made it easier to detect the social challenges of boys with autism spectrum disorder. Typically
developing boys tended to play organized games; boys with autism spectrum disorder tended to play alone. The results
highlight a male bias in our perception of autism spectrum disorder. If practitioners look for social isolation on the
playground when identifying children with social challenges, then our findings suggest that girls with autism spectrum
disorder will continue to be left unidentified.

Keywords
autism spectrum disorder, gender, girls, school, sex, sex differences, social behaviors

Introduction
Although much progress has been made in understanding Given that children with ASD without cognitive impair-
autism spectrum disorder (ASD), researchers have histori- ment are often educated in the general education setting,
cally experienced difficulties in identifying and diagnos- with typically developing (TD) children prominent in the
ing girls with ASD without cognitive impairment (Shattuck social landscape, schools provide a relevant context to
et al., 2009). Empirical findings suggest that this gap may examine sex and gender in ASD populations. Sex refers to
be due, in part, to an expectation that girls with ASD will male or female biological traits. Gender describes psycho-
appear as solitary on the playground as boys with ASD do logical characteristics and social behaviors that are per-
(Hiller et al., 2014)—that is, there is a male bias to our ceived as either masculine or feminine (Jordan-Young,
expectations. Concomitantly, girls are described as being 2010). Sex is a powerful predictor of children’s social
better able than boys to “camouflage” their symptoms of organization; TD boys and girls naturally segregate into
ASD and to use compensatory behaviors that mitigate their
social challenges (Dworzynski et al., 2012; Gould and 1CaliforniaState University Channel Islands, USA
Ashton-Smith, 2011; Tierney et al., 2016). The word cam- 2Maternal and Child Health Bureau, MD, USA
ouflage highlights the importance of the environment. 3University of California, Los Angeles, USA

Looking more closely at the way that boys and girls with
Corresponding author:
ASD interact with, or blend into, their natural social envi- Michelle Dean, California State University Channel Islands, 2602
ronment at school may give us a better understanding of Madera Hall, One University Drive, Los Angeles, CA 93012, USA.
why it has been difficult to identify ASD in girls. Email: [email protected]
Dean et al. 679

same-sex peer groups (Fabes et al., 2003; Maccoby, 1999; because teachers are more likely to notice and report con-
Mehta and Strough, 2009), and qualitative differences in cern for overt behaviors (Dworzynski et al., 2012; Hiller
the ways that boys and girls socialize are widely recognized et al., 2014; Mandy et al., 2012), a boy is more likely to be
(Corsaro and Eder, 1990; Goodwin, 2006; Pellegrini et al., identified and to receive intervention services.
2004). Gender identity becomes more pronounced in Similar findings have been reported in the special edu-
school environments, especially within same-sex peer cation literature, which identified sex as a salient predictor
groups (Maccoby, 1988). Similar to TD children, children for special education referral and subsequent diagnoses
with ASD also tend to segregate by sex when socializing (McIntyre and Tong, 1998; Gregory, 1977). Because boys
at school (Dean et al., 2014). More research is needed to are more likely to manifest externalizing and disruptive
examine the extent to which gender-related factors influ- behaviors, they are significantly more likely to be referred
ence the social acceptance of children with ASD at school. for special education evaluation (Dhuey and Lipscomb,
Although most children in elementary school socialize 2010). Consequently, girls with disabilities are at risk for
on the playground during break time or recess (Blatchford having unmet service needs (Bussing et al., 1998) and neg-
et al., 2003), boys and girls tend to interact differently. TD ative outcomes (Arms et al., 2008). Girls with ASD appear
boys tend to play in large stable groups; they socialize to be following a similar trend. The inability of school per-
through activities (Corsaro and Eder, 1990), preferring sonnel to recognize symptoms of ASD has been identified
rough-and-tumble (Corsaro and Elder, 1990; Maccoby, as a significant factor in the misdiagnosis and late diagno-
1988), organized play, and competitive-team games sis of children with ASD. Compared to boys with ASD,
(Pellegrini et al., 2004). Female groups are smaller and ASD symptomology is significantly less likely to be rec-
more exclusive than male groups (Goodwin, 2006). Girls ognized in girls (Aggarwal and Angus, 2015). Using a
spend more time in social conversation and build intimacy population-based sample of boys and girls who presented
through sharing secrets (Maccoby, 2004). Within a gender with high levels of ASD traits, Dworzynski et al. (2012)
socialization framework, certain ASD-related behaviors examined differences between children who received an
may be differentially reinforced or sanctioned in male ASD diagnosis and children who did not meet the diagnos-
versus female groups (Tierney et al., 2016). For example, tic criteria. Although ASD trait levels were similar, girls
social communication difficulties may interfere with receiving an ASD diagnosis had an additional low IQ or
engaging in intimate conversations, but not with playing teacher-reported behavior problems. These differences
kickball. Likewise, certain social behaviors, such as were not evident in the equivalent group of boys. The find-
playing team sports, may be of more value in male ings suggest that the current diagnostic criteria and proce-
groups compared to female groups (Goodwin, 2006). dures may overlook the female endorsement of ASD
Consequently, our understanding of how gender influ- symptomatology. The behaviors of girls with ASD, who
ences ASD-related behavior necessitates an examination are more prone to internalizing symptomatology (Mandy
of the reciprocal, interactive relationships between chil- et al., 2012; Solomon et al., 2012), are more likely to be
dren and their social environment. tolerated by their teachers with no report of concern
Research suggests that socially constructed gender (Dhuey and Lipscomb, 2010; Dworzynski et al., 2012;
biases shape the way that ASD-related behaviors are toler- Hiller, 2016; Mandy et al., 2012).
ated and perceived (Lai and Baron-Cohen, 2015). Clinical Relative to boys with ASD, girls with ASD are described
descriptions and research findings have identified gender as having superior interpersonal skills (Hiller et al., 2014;
differences in the way that boys and girls endorse ASD- Lai and Baron-Cohen, 2015; Mandy et al., 2014). Clinical
related behaviors. Reported differences indicate that it reports describe girls with ASD as mimicking the social
may be easier to detect ASD behaviors in boys. For exam- behaviors of others and appearing able from an observer’s
ple, research findings suggest that boys with ASD are perspective to mask their social deficits, repetitive behav-
prone to isolation (Hiller et al., 2014) and have signifi- iors, and restricted interests (Attwood, 2006; Kopp and
cantly more restrictive interests and repetitive behaviors Gillberg, 1992). In an empirical study, Hiller et al. (2014)
than girls (Hartley and Sikora, 2009; Hiller et al., 2014; examined caregiver reports about the ASD-related behav-
Lord, 1982; Mandy et al., 2012). Likewise, boys with ASD iors their children endorsed (n = 152; 60 were completed
are more likely to show repetitive interests that strike oth- for females). Findings indicated that being female was a
ers as unusual (e.g. hand flapping and obsession with train significant predictor of complex imitation skills, with par-
schedules), whereas girls with ASD often show interests ents describing girls as being better able than boys to copy
that seem more in line with the interests of TD children the social behaviors of their peers. Yet, despite their rela-
(Gould and Ashton-Smith, 2011; Hiller et al., 2014; Kopp tive strengths and use of compensatory behaviors, recent
and Gillberg, 1992). Research findings also indicate that research examined social networks of school-aged chil-
boys with ASD exhibit greater externalizing symptomol- dren with and without ASD (Dean et al., 2014) and found
ogy, hyperactivity, and inattention compared to girls with that girls with ASD experience social challenges that are
ASD (Gould and Ashton-Smith, 2011; Hiller et al., 2014); similar to boys with ASD vis-à-vis social acceptance,
680 Autism 21(6)

Table 1. Descriptive statistics of ASD sample.

Boys, µ (σ) Girls, µ (σ) t (df) Sig.

Age (years) 7.71 (1.23) 7.75 (1.22) 0.12 (46) 0.91
Grade 2.29 (1.08) 2.50 (1.25) 0.62 (46) 0.54
ADOS3 SA 11.46 (4.12) 10.13 (4.59) −0.81 (46) 0.42
ADOS3 RRB 3.13 (2.15) 2.33 (1.34) −0.80 (46) 0.43
ADOS3 OT 14.58 (5.56) 12.46 (5.06) −0.93 (46) 0.36
ADOS3 Severity 7.83 (2.26) 7.08 (2.28) −1.00 (46) 0.32
Abbreviated SB-5 94.50 (12.31) 93.50 (14.36) 0.51 (46) 0.61

df: degree of freedom; ADOS3: Autism Diagnostic Observation Schedule Module 3; SA: Social Affect; RRB: Restricted and Repetitive Behaviors; OT:
Overall Total; SB-5: Stanford–Binet Intelligence Scale: Fifth Edition.

reciprocal friendships, and social network status. Gender Sample

differences were evident in the way that children with
ASD were rejected, thus highlighting a potential bias in ASD sample. Children with ASD had a confirmed diagnosis
our ability to recognize the social challenges of girls. Boys of ASD (Autism Diagnostic Observation Schedule
with ASD were overtly rejected, while girls with ASD (ADOS); Lord et al., 2002) without intellectual disability
were neither accepted nor rejected; instead, they appeared (IQ ⩾ 70; Abbreviated Stanford–Binet Fifth Edition) and
to be overlooked and neglected (Asher et al., 2001; Dean were educated in the general education classroom for a
et al., 2014). These findings suggest that there are potential minimum of 80% of the school day. In the original study
benefits to identifying the social behaviors of girls and (Kasari et al., 2015), the ratio of male participants to female
boys with ASD within their natural social environment at participants followed epidemiological reports (4–8:1);
school; moreover, research is needed to examine the extent therefore, data from all the girls with ASD were selected
to which gender facilitates social involvement. (n = 24). The male sample (n = 24) was randomly selected
The purpose of this study is to examine the following from the pool of participants that matched girls with ASD
research questions. (1) To what extent do environmental by age, grade, IQ, and school. Random selection of the
factors such as gender-related social behaviors and activities male group controlled for the potentially confounding
play a role in helping girls with ASD to mask their symp- effects of school and site differences. The average IQ of the
toms? (2) Are girls with ASD better at “camouflaging” their participants with ASD was 94.00 (13.24), which is slightly
symptoms of ASD and using compensatory behaviors to lower than would be expected in TD populations. Yet,
mitigate their social difficulties? (3) Are the symptoms of all participants in this sample were fully included in the
ASD more obvious and easier to detect in boys? general education classroom, so despite the potentially
confounding effects of IQ, this sample is representative of
children with ASD who are educated in general education
Methods settings. There were no significant differences between
boys and girls on age, grade, cognitive abilities, or ASD
Data source symptomology as measured by the ADOS (Table 1).
This is a secondary analysis of data drawn from an earlier
study (Kasari et al., 2015), a large multisite randomized TD control group. Peers were randomly selected from a
control trial that compared two types of social skills’ inter- pool of participants matching children with ASD by sex/
ventions at schools. Participants were elementary school- gender, grade, age, and city of residence. The results of a
aged children in first through fifth grades with and without 2 × 2 (sex × diagnosis) analysis of variance (ANOVA) test
ASD who lived in Los Angeles, Baltimore, Seattle, and indicate that the effects of age (µ = 7.92, σ = 1.22) and
Ann Arbor. After meeting criteria to participate in the study, grade (µ = 2.8, σ = 1.30) were not significant (Table 2).
children with ASD were randomized to either the control
condition (pull-out social skills’ group) or an experimental
Measures
condition (peer-mediated social skills’ group). TD children
with positive social skills (per teacher nomination) were Eligibility criteria
recruited for the peer-mediated condition. There were two
to three TD peers for every one child with ASD in the peer- ASD. The ADOS (Lord et al., 2002) was used to confirm
mediated condition. All participants signed written assent diagnosis of ASD. It is a standardized, semi-structured
and had written parent consent. The primary outcome data play-based assessment of autistic symptoms. Based on
used in this study were collected prior to the start of the expressive language ability (Lord et al., 2000), PhD stu-
intervention. See Kasari et al. (2015) for more information dent researchers and post-doctoral scholars who were
about the original data collection procedures. trained and reliable administered the ADOS Module 3 to
Dean et al. 681

Table 2. Results of a 2 × 2 (sex × diagnosis) ANOVA on age and grade.

Role Sex Sex × role

F (df) Sig. ηp F (df) Sig. ηp F (df) Sig. ηp

Age (years) 0.50 (3, 92) 0.48 0.01 0.01 (3, 92) 0.94 0.00 0.06 (3, 92) 0.81 0.01
Grade 2.43 (3, 92) 0.12 0.34 0.01 (3, 92) 0.94 0.05 0.82 (3, 92) 0.37 0.15

df: degree of freedom.

all participants. Using the ADOS 2 revised algorithm, Because of widely reported gender differences in the way
scores were converted to the domains of Social Affect that boys and girls play, we examined the Game and Joint
(SA), Restricted and Repetitive Behaviors (RRB), Overall Engagement variables separately.
Total (OT), and Overall Severity (OS). Qualitative descriptions of the students’ social behavior
were also recorded during each 1-min interval. After clas-
Cognitive ability. The Stanford–Binet Intelligence Scale: sifying the participant’s engagement state, the observers
Fifth Edition (SB-5) is a valid and reliable standardized recorded in an open-ended field note, focusing on the sub-
test that measures intelligence and cognitive abilities in ject’s peer group, the presence or absence of a conversa-
children and adults. We used the abbreviated version, tion, the social activity, and the child’s affect. An example
yielding a non-verbal and verbal IQ score. The abbreviated of a POPE observation is shown in Table 3.
IQ scores are highly correlated with full-scale IQ scores.
Analysis
Primary outcome variables We used concurrent mixed methods (QUAN + QUAL), in
Playground Observation of Peer Engagement. The Playground which quantitative and qualitative data were collected at
Observation of Peer Engagement (POPE) is a timed inter- the same time and analyzed concurrently. Both parametric
val behavior coding system that yields quantitative and and non-parametric tests were used since not all quantita-
qualitative data and has been successfully applied on public tive data were normally distributed. The results of the non-
school playgrounds in various peer observation studies parametric tests and parametric tests were statistically
(Kasari et al., 2011, 2015; Kasari and Rotheram-Fuller, similar. Because descriptive statistics will show the mean
2005). Independent and blinded observers watched chil- scores and standard deviations, the final results are based
dren with and without ASD on the playground during on the parametric tests for ease of interpretation.
recess for a minimum of 10 min and a maximum of 15 min. For the qualitative analysis, we used an exploratory
Engagement states and behaviors were recorded in 1-min case study design, with each group representing one case
intervals; an observer watched the target child for 40 s and (Girls ASD, TD Girls, Boys ASD, TD Boys). First, we cat-
coded for 20 s. Observers were trained to a reliability crite- egorized the field notes by engagement state (Game, Joint
rion of α > 0.80. Two raters randomly and independently Engage, and Solitary). Next, we used line-by-line coding
coded 20% of the observations, maintaining an average at the individual level to identify participant activities dur-
inter-rater reliability of 0.87 (see Kasari et al., 2011, 2015 ing each 1-min interval. Coders were the first author and
for more information on observer training procedures). two student research assistants. All coders were blind to
Three engagement states are the focus of this study: sex, age, grade, and diagnosis during coding procedures.
Game (the child is actively playing a game with a peer or Twenty-five percent of the data were double coded, and
peers), Joint Engagement (the child is actively socializing coding consistency was above 0.90. Activities listed most
with a peer or peers), and Solitary (the child is alone and frequently during the observation were labeled “primary
not engaging with other children). Engagement states activities,” such that each participant had primary activi-
were recorded during each 1-min coding interval, and the ties in Game, Joint Engage, and Solitary. For example, if a
engagement state variables describe the proportion of the participant talked for 8 min and tossed a football for 2 min,
observation period that the participant spent in each state. the primary Joint Engage activity would be talking. The
When the Engagement states of children with ASD engagement state data were separated by case after the first
have been reported in prior studies, the Game and Joint phase of coding. To determine primary activities at the
Engagement variables were combined to create a “Joint group level, we counted the frequency that each primary
Engage” aggregate variable, and the aggregate variable activity occurred within each group and engagement state.
described the amount of time children were mutually To be considered in the comparison, more than one child in
involved in activities with peers, without discriminating the entire sample had to engage in the activity for more
between structured games with rules and socializing with- than a 1-min interval. Finally, we compared primary activ-
out structured rules (Kasari, 2015; Kasari et al., 2011). ities across cases.
682 Autism 21(6)

Table 3. Example of POPE variables and observational field notes.

Minute Engagement state Description

Solitary
1 S Wandering the yard by himself
2 S Sits next to a paraprofessional, conversation
3 S Talks to paraprofessional, neutral affect
4 S Walking around the yard, neutral affect
Joint Engage
1 JE Walking and talking with two friends, laughing
2 JE Sits with two friends, talking and sharing snack
3 JE Sharing snack, new friends join, happy
4 JE Sitting and talking to one friend, serious
Game
1 G Playing handball with a group of boys
2 G Got out, sits on bench and watches the game
3 G Sitting on bench waiting turn
4 G Has a turn, in the game, serious

S: Solitary; JE: Joint Engagement; G: Game.

Table 4. Descriptive statistics and significant between-group differences on Game, Joint Engage, and Solitary.

TD boys TD girls ASD boys ASD girls Significant between-group differences

Mean (SD) Mean (SD) Mean (SD) Mean (SD)

Games 41.50 (35.91) 13.75 (27.84) 10.87 (18.63) 6.68 (12.63) TD (boys and girls) > ASD (boys and girls)
Boys (TD and ASD) > girls (TD and ASD)
TD boys > TD girls, ASD boys, ASD girls
Joint Engage 31.67 (31.44) 52.08 (35.01) 23.55 (27.80) 39.00 (31.46) Girls (TD and ASD) > boys (TD and ASD)
Solitary 3.81 (6.91) 7.92 (14.46) 43.57 (33.90) 26.69 (28.51) ASD (boys and girls) > TD (boys and girls)
Boys ASD > TD girls, TD boys, ASD girls

TD: typically developing; ASD: autism spectrum disorder; SD: standard deviation.

Preliminary quantitative and qualitative results deline- interaction effect on Games indicated that TD boys (F(3,
ated the salient engagement states and the primary activities 92) = 5.18, p = 0.025, ω2 = 0.053) spent significantly more
of each group. We used the findings to create a social profile time playing Games than all other groups.
for each group. With the exception of TD girls playing
games, children’s social behaviors were consistent across
Joint Engage
grade levels. Finally, representative examples from the raw
data were selected and used to illustrate the social profiles. There was a significant effect of sex/gender on Joint
Engage, in which girls with and without ASD spent more
time in Joint Engage than boys with and without ASD
Results (F(3, 92) = 7.76, p = 0.006, ω2 = 0.078). Role and the inter-
Descriptive statistics of the primary outcome variables are action were not significant (role: F(3, 92) = 2.71, p = 0.103,
shown in Table 4. Using a 2 × 2 ANOVA, we identified ω2 = 0.029; interaction: F(3, 92) = 0.149, p = 0.701,
significant between-group differences in Game, Joint ω2 = 0.002).
Engage, and Solitary.
Solitary
Game There was a significant effect of role and the interaction
There was a significant effect of sex/gender, role, and on Solitary. Children with ASD spent significantly more
interaction on Game. Boys play games significantly more time in Solitary than TD children (F(3, 92) = 37.04,
than girls (F(3, 92) = 9.52, p = 0.003, ω2 = 0.094), and TD p = 0.000, ω2 = 0.287). The interaction effect shows that
children play games significantly more than children with boys with ASD spent significantly more time in Solitary
ASD (F(3, 92) = 13.27, p = 0.000, ω2 = 0.126). The than all other groups (F(3, 92) = 4.37, p = 0.03, ω2 = 0.05).
Dean et al. 683

Table 5. The number of participants within each group (n) and the proportion of the total observation intervals (%) that
participants engaged in each activity.

TD ASD

Boys Girls Boys Girls

n % n % n % n %
Game activities
Pretend play – – – – 1 33 1 40
Tag 2 75 1 40 1 20 3 13
Jump rope game 1 17 1 13 1 73 1 77
Card game 2 40 – – – – – –
Board game 2 75 – – – – – –
Team ball games 6 67 – – 2 29 – –
Single-player ball games 2 52 4 74 2 60 – –
Table top games 4 57.50 – – – – – –
Total 19 6 6 4
Average 52 27 42 43
Joint Engage activities
Talking 7 41 8 54 5 17 11 44
Snow play 1 77 – – – – – –
Setting up game 1 50 – – – – – –
Walking 2 12 – – – – – –
Play structure 2 35 1 70 3 20 1 20
Chase 1 27 – – – – – –
Lining up to go to class 1 30 – – – – – –
Drawing – – 2 71 – – – –
Jump rope – – 3 32 – – 1 18
Flittinga – – 5 100 2 77 2 60
Sharing snack – – 2 63 – –
Pretend play – – – – 1 27 2 33
Unstructured ball game – – – – 2 42 1 9
Digging in sand/dirt – – – – 1 27 1 27
Total 15 21 14 19
Average 39 65 35 30
Solitary activities
Flita – – – – – – 2 60
Looking/waiting for friend – – – – – – 2 13
Eating – – – – – – 1 20
Wandering – – – – 5 41 2 20
Play structure – – – – 3 39 1 70
Kickball – – – – – – 1 56
Sandbox – – – – – – 1 60
Adult – – 1 13 2 37 1 6
Snow – – – – – – 1 75
Running 1 10 – – 2 15 – –
Repetitive behavior – – – – 1 18 – –
Basketball – – – – 2 58 – –
Laying down – – – – 1 27 – –
Jump rope – – – – 1 36 – –
Sitting – – – 18 – – – –
Drawing – – – 15 – – – –
Total 1 4 17 12
Average 10 15 34 42

TD: typically developing; ASD: autism spectrum disorder.

aFlitting: when children spent about the same amount of time in three or more activities during the observation period.

The effect of sex was not significant (F(3, 92) = 1.76, Table 5 denotes the primary activities of each case in
p = 0.187, ω2 = 0.019). Game, Joint Engage, and Solitary.
684 Autism 21(6)

Social profiles boys exhibited solitary behavior (µ = 3.81, σ = 6.91); one

boy ran alone for 1 min of his observation.
TD boys. TD boys spent a majority of recess in Game
The representative examples below highlight the char-
(µ = 41.50, σ = 35.91), and the significant interaction effect
acteristics of TD boys’ social behaviors. Both examples
on Game indicated that TD boys spent more time in Game
demonstrate the physical nature of boys’ play. In the first
than all other groups. TD boys also spent a large amount of
example, the TD boy interacts with peers while playing a
time in Joint Engage (µ = 31.67, σ = 31.44). The most sali-
structured game with rules throughout the observation.
ent primary activities were playing team ball games, which
The second example highlights the physical nature of play
boys played for an average of 67% of the observation, and
during Joint Engage.
talking, for an average of 41% of the observation. Few TD

State Field note

TD boys—Observation 1
Game Playing handball, got out talked to girl on bench, boy came over conversation.
Game In game, get interesting responses to friendly chuckles from boys on bench, laughing.
Game Laughing, playing handball, friendly competition, everyone laughing.
Game Conversation on bench, watching and laughing play. Conversation with people on
bench.
Game Waiting turn on bench, eating grapes, gets to play for 40 seconds. He is happy.
Game Call out at play, sharing grapes with friend on bench, watching game. He is happy.
TD boys—Observation 2
Joint Engage Resting, then jumped up to be chased by all three other boys. Laughing.
Joint Engage Fell during chase, talking and holding self in exaggerated pain (not serious).
Joint Engage Sitting out of chase for a minute. A couple of exchanges with peers.
Joint Engage Ran away again suddenly. Laughing and smiling.
Joint Engage Rest again in the shade. Talking to two buddies.
Joint Engage Walking with friends toward the yard.

Each line denotes 1 min of observation.

TD girls. Compared to TD boys, TD girls were less inclined activities throughout the observation. TD girls spent very
to play games during recess. However, when they did play little time in Solitary (µ = 7.92, σ = 14.46), with four TD
games, TD girls tended to be older (grades 4 and 5) and to girls exhibiting primary activities in Solitary, lasting, on
play single-player ball games (e.g. tetherball and four average, 15% of the observation period.
square), rather than team ball games. TD girls spent a The representative examples highlight the fluidity in
majority of recess in Joint Engage (µ = 52.08, σ = 35.01), the social behaviors of TD girls. In the first example, the
significantly more than boys with and without ASD, but TD girl moves through multiple activities while talking in
not statistically more than girls with ASD. The most popu- a small group, and maintaining Joint Engagement. The
lar primary activities of TD girls were talking (averaging second example is representative of fluidity occurring in
54% of the observation) and flitting (averaging 100% of the Game engagement state. In this example, the TD girl
the observation). An activity was labeled “flitting” when a moves back and forth between groups while maintaining
child spent equal amounts of time in three or more Game status throughout the duration of the observation.

Engagement state Field note

TD girls–Observation 1
Joint Engage Walking around with 2 other girls, smiling, happy.
Joint Engage Walking around, arm in arm. Smiling, laughing.
Joint Engage Sitting in a circle. Talking and pretending to paint each others’ nails
Joint Engage Still pretending to paint nails, humming, happy
Joint Engage The 3 girls get up. She grabs one by the hand and they run off :)
Joint Engage They start chasing a boy around, all laughing and happy.
TD girls–Observation 2
Game Four square.
Game Briefly joined boys “hot potato” game nearby, then returned to four square.
Game Four square.
Game Joined other group’s game.
Game Hot potato game.

Each line denotes 1 min of observation.

Dean et al. 685

Boys ASD. Boys with ASD spent significantly less time 41% of the observation and talking for an average of 17%
than TD boys in Game (µ = 10.87, σ = 16.63). Although of the observation.
they were less involved, six boys with ASD participated The following examples are representative of the social
in some structured games with activities for an average of behaviors of boys with ASD during recess. In the first
27% of the observation period. Similar to TD boys, boys example, the boy with ASD is in Solitary engagement
with ASD also spent a notable amount of time in Joint throughout the entire observation period. The second
Engage (µ = 23.55, σ = 27.80). Solitary was the most sali- example is of a boy with ASD who successfully engaged
ent engagement state for boys with ASD, who spent sig- in a Game throughout the duration of recess. His ASD-
nificantly more time in Solitary than all other groups related behaviors are obvious; despite his successful
(µ = 43.57, σ = 33.90). The most popular primary activi- engagement, the boy uses repetitive behaviors throughout
ties of boys with ASD were wandering for an average of the observation.

Engagement state Field note

Boys with ASD Observation 1
Solitary Sitting at table with two aides. No other kids around.
Solitary Smiling and staring at juice box, starts to go to yard.
Solitary Wandering around with head down. Alone, looking at his shadow.
Solitary Aide explains that he can’t look at his shadow while he’s walking. Only when he stops so he
doesn’t run into anyone.
Solitary Still wandering around alone. Neutral affect.
Solitary Digging hole in the dirt with his foot. Alone. Neutral affect.
Boys with ASD Observation 2
Game Positive affect. Joins game and takes turn. Gets called out. Sits out and calls another kid out.
Game Watching game. Repetitive behavior and covering ears. Girl initiates and he covers ears.
Game Watching and hand mannerisms. Repetitive behavior. Positive Affect.
Game Rocking and clapping hands. Seems excited and anxious for his turn.
Game Watching intently. Kicking legs. Takes his turn and gets called out.
Game Girl sits right next to him and smiles. No response. Waiting for turn.

Each line denotes 1 min of observation.

Girls ASD. Girls with ASD spent little time in Game (n = 4; with and without ASD at school. Gender differences in
averaging 47% of the observation). Similar to TD girls, social behaviors were evident throughout the sample—TD
girls with ASD spent most of the time in Joint Engage boys tended to play differently than TD girls, and the social
(µ = 39.00, σ = 31.46), and talking was a top primary activ- challenges of boys with ASD looked different, and were
ity for girls with ASD (for an average of 44% of the obser- more obvious, than the social challenges of girls with
vation). Unlike TD girls, girls with ASD also spent a ASD. Gender-related social norms were evident in the
significant amount of time in Solitary (µ = 26.69, σ = 28.51). social landscape. From the perspective of untrained
Their second primary activity was flitting, which appeared observers, such as recess aides and many teachers, male
in Joint Engage and in Solitary for an average of 60% of social groups were as conducive to exposing the social
the observation in either engagement state. challenges of boys, as female groups were to camouflag-
Unlike TD girls, who are readily accepted into activities ing girls’ social challenges.
with peers, girls with ASD appear to use compensatory The female social landscape supports the camouflage
behaviors to gain access into peer groups that may mask hypothesis; the fluidity of female social groups created an
their social challenges. In the first example, the girl with ideal backdrop to conceal the girls with ASD who were
ASD weaves in and out of Joint Engage and Solitary, high- often hovering close by. Regardless of engagement state
lighting her difficulties maintaining mutual involvement in status, girls with ASD tended to stay in close proximity to
social groups. The girl with ASD in the second example social groups and were therefore better situated than boys
maintains Game by swinging a jump rope throughout the with ASD to capitalize on social opportunities. Yet, this
observation period. Social challenges are evident, how- relative strength has some iatrogenic effects. Scanning the
ever, because she is never given a turn to jump. playground environment, as one would expect a play-
ground attendant to do, would be insufficient to identify
the social struggles of girls with ASD. From a distance,
Discussion girls with ASD looked like TD girls. They spent a signifi-
We used mixed methods to examine playground observa- cant amount of time talking in Joint Engage and weaving
tions of the social engagement and activities of children in and out of groups. Girls with ASD, however, tended to
686 Autism 21(6)

Engagement state Field note

Girls with ASD Observation 1
Joint Engage Caterpillar activity four girls total. She is jumping in the letters (painted in each circle). Neutral Affect.
Solitary Walked away from the girls, near a crowd of boys who are standing around. No talking. Neutral.
Solitary Walking near a group, but not part of the group. There are three girls with each other, and the target is
standing nearby, but by herself.
Solitary Standing near 2 girls looking around. She is not part of a group, but close. Neutral Affect.
Joint Engage Three starts to try to initiate to the girls to do something else. They have a conversation.
Joint Engage Walking on the yard with girls. She has one conversation. The girls try to walk away and she goes with them.
Girls with ASD Observation 2
Game Playing jump rope with younger kids, swinging the rope, but not talking or interacting.
Game Same, swinging the jump rope as the other girls are jumping, not talking to them.
Game Talking briefly with girls about the game. Then the other girl leaves and she observes the others playing jump
rope.
Game Swinging jump rope for girls.
Game Swinging jump rope for girls.
Swinging the jump rope with another girl again. She is always swinging the rope, but never in the middle
jumping.

Each line denotes 1 min of observation.

flit in between Joint Engage and Solitary; TD girls main- scanning the playground environment on a regular basis
tained their Joint Engage status while moving from one would be sufficient to identify the social struggles of boys.
group to the next. This suggests that while the girls’ social Therefore, the differential recognition of the social chal-
challenges are concealed from playground attendants, they lenges girls and boys with ASD face at school highlights a
do not appear to be hidden from peers. male bias in our perceptions of children’s peer
It is the significant proportion of time spent in Solitary, interactions.
regardless of proximity, that marks social withdrawal Interestingly, our findings suggest that children with
or exclusion. Not being able to maintain mutual engage- ASD showed some sensitivity to socially constructed gen-
ment in activities highlights girls’ difficulties with social der expectations. In general, girls with ASD engaged in
synchronization. TD girls use subtle sanctions to enforce unstructured social activities (Joint Engage) that were sim-
group norms and shape social violations (Goodwin, ilar to TD girls (e.g. talking, flitting), while the proportion
2006). Given the social deficits related to their diagnosis, of time that boys with ASD spent in Joint Engagement was
it may be difficult for girls with ASD to recognize and about the same as TD boys. Compared to girls with ASD,
interpret social cues (e.g. eye roll, shared glance, giggle, boys with ASD played more structured games with rules,
or smirk; Card et al., 2008) and to adjust their own behav- suggesting that the disproportionate amount of time that
ior in order to align with group norms (Dean et al., 2013). boys with ASD spent in Solitary may be more indicative of
This lack of ability to conform can mistakenly be inter- deficits related to ASD (e.g. psychomotor coordination, or
preted as blatant social violations, putting girls with turn taking challenges), than of a violation of gender
ASD at risk for relational aggression and exclusion norms. Consequently, physically demanding play may be
(Crick and Dodge, 1996; Dean et al., 2013; Goodwin, prohibitive for some boys with ASD. Although gender-
2006). Consequently, using camouflage to mask social related trends in play preferences were evident, we were
challenges makes girls vulnerable and less likely to not able to determine whether the children with ASD
receive intervention. received social reinforcement by selecting gender-appro-
The male social landscape makes it easier to detect the priate social activities.
social challenges of boys with ASD. A large proportion of These results suggest that it is necessary to adjust pro-
TD boys played structured games with rules or engaged in tocols to reduce gender bias when working with ASD chil-
physically active play. Given the stability and prevalence dren in educational settings. Gender differences were
of games (Pellegrini et al., 2004), it was relatively easy to evident in the way that children without ASD socialized,
spot the boys with ASD, who were often wandering alone and because of this, girls with ASD likely have different
apart from the game. Given the significant proportion of social experiences than boys with ASD. Considering gen-
time spent in Solitary—a stark contrast to TD boys—boys der differences in play styles and weighing the social land-
with ASD appeared to have difficulty accessing and inde- scape at school may help reduce the male bias in our ability
pendently maintaining involvement in male play groups. to identify girls with ASD at school. Understanding how
So contrary to the social challenges of girls with ASD, children with ASD fit into the social landscape at school is
Dean et al. 687

a necessary step when developing social interventions. lacks skills to engage, versus being solitary because one
Additionally, given the male-to-female ratio among chil- enjoys time alone. More research is needed to examine the
dren identified with ASD, it is common for one girl to underlying motivations that drive the Solitary engagement
receive social skills interventions with a group of boys, of children with ASD at school. Fifth, the TD sample was
which may limit girls’ ability to develop social skills nominated to participate in the original study because of
needed to be successful in female groups (Cridland et al., their prosocial behaviors. More research is needed to
2014). The short, interrupted periods of engagement sug- examine the social behaviors and social acceptance of chil-
gest that girls would benefit from working on the quality dren with ASD compared to TD children with varying
of their interactions, to learn how to interpret social cues degrees of social competence. Finally, although our study
and to adjust their behavior accordingly. provides insight into differences in the way that boys and
In addition to further work in reducing gender bias, girls with ASD socialize, more research is needed to exam-
more research is needed to develop and test social inter- ine the quality of social interactions, to identify where
ventions that focus on female social skills, to better under- breakdowns occur, and to design and implement interven-
stand communication breakdowns. Given the relative tions within the children’s real-life school culture.
strengths of girls (e.g. the ability to gain access into peer
groups), girls with ASD may benefit from strength-based
approaches to support their participation in peer groups Conclusion
(Wilding and Griffey, 2015). The male environment was This study supports the camouflage hypothesis and
more physically demanding than the female environment, increases our understanding of the social behaviors of girls
suggesting that boys with ASD might be more successful if and boys with ASD in the natural social setting at school.
schools facilitated structured play activities without heavy Past research has described a relative strength in girls with
physical demands. ASD, who are better able than boys to mask their symp-
toms from adult observers. Being in close proximity to
Limitations peer groups helped girls to access opportunities for social
interaction, but a closer look suggests that girls with ASD
There are limitations related to the size and scope of this were less likely to have the skills necessary to use those
study. First, we used one observation to assess the play- opportunities to engage successfully with peers. If practi-
ground behavior of children at school. Longitudinal tioners continue to tolerate or overlook subtle social chal-
research is needed, using longer observation periods across lenges and internalizing behaviors, then camouflage is not
multiple days and time points over the course of a school protective. In contrast, the conspicuousness of ASD in
year. If girls with ASD have difficulty synchronizing with boys makes it easier for boys to get access to intervention.
group norms, it is conceivable that time spent in Solitary Compensatory behaviors, therefore, are advantageous
will increase over the course of a school year. Second, col- only to the extent to which the needs of children who
lecting observation information in conjunction with cogni- engage in such behaviors are met, either through mutual
tive measurement and parent, peer, and teacher reports of engagement with peers or through intervention support.
behaviors may build a more comprehensive understanding
of gender differences in the behaviors of children with Acknowledgements
ASD at school and deepen our understanding about how
The Autism Intervention Research Network for Behavioral
girls and boys with ASD interpret social contexts and
Health team includes Connie Kasari and Fred Frankel at Center
interactions. Moreover, more research is needed to exam- for Autism Research and Treatment, UCLA; Rebecca Landa at
ine the overlay of environmental expectations on biologi- Kennedy Krieger Institute, Johns Hopkins University School of
cal traits. Examining children’s play preferences and social Medicine, Baltimore, MD; Catherine Lord at Center for Autism
behaviors through gender socialization theory concomi- and the Developing Brain, Weill Cornell Medical College; Felice
tantly with a biological lens (Knickmeyer et al., 2007) can Orlich and Brian King at Seattle Children’s Hospital, University
increase our understanding of children’s social synchroni- of Washington; Robin Harwood at the U.S. Maternal and Child
zation and the extent to which social acceptance is related Health Bureau. The information or content and conclusions are
to innate preferences (and therefore shared interests) ver- those of the author and should not be construed as the official
sus social shaping through environmental stimuli. Third, position or policy of, nor should any endorsements be inferred by
although we were able to make equal group size compari- HRSA, HHS, or the US Government. The data used in this study
were collected as a part of a larger study, Children with ASD and
sons, our sample was relatively small, and we did not have
social skills groups at school: Randomized trial comparing inter-
enough statistical power to perform a cross-sectional anal- vention approach and peer composition, which was published in
ysis of gender-related behaviors across a developmental the Journal of Child Psychology and Psychiatry in 2015. The
span. More research is needed to examine the social authors received Institutional Review Board’s Approval at the
behaviors and acceptance of boys and girls with ASD University of California, Los Angeles, the University of
throughout elementary and secondary school. Fourth, Michigan, John Hopkins, the University of Washington, and each
there is a difference between being solitary because one of the local school districts from which the data were collected.
688 Autism 21(6)

All participants had written parent consent and signed assent Fabes RA, Martin CL, Hanish LD, Anders MC and Madden-
forms prior to participating in the study. This article examines Derdich DA (2003) Early school competence: the roles of
observation data and focuses specifically on the social behaviors sex-segregated play and effortful control. Developmental
of girls and boys with and without ASD during recess. Psychology 39(5): 848.
Goodwin MH (2006) The Hidden Life of Girls: Games of Stance,
Funding Status, and Exclusion. Malden, MA: Blackwell.
Gould J and Ashton-Smith J (2011) Missed diagnosis or misdi-
This project was supported by the Health Resources and Services
agnosis? Girls and women on the autism spectrum. Good
Administration (HRSA) of the US Department of Health and
Autism Practice (GAP) 12: 34–41.
Human Services (HHS) under grant no. UA3MC11055, Autism
Gregory MK (1977) Sex bias in school referrals. Journal of
Intervention Research Network on Behavioral Health.
School Psychology 15(1): 5–8.
Hartley SL and Sikora DM (2009) Sex differences in autism
References spectrum disorder: an examination of developmental func-
Aggarwal S and Angus B (2015) Misdiagnosis versus missed tioning, autistic symptoms, and coexisting behavior prob-
diagnosis: diagnosing autism spectrum disorder in adoles- lems in toddlers. Journal of Autism and Developmental
cents. Australasian Psychiatry 23: 120–123. Disorders 39: 1715–1722.
Arms E, Bickett J and Graf V (2008) Gender bias and imbal- Hiller RM, Young RL and Weber N (2014) Sex differences in
ance: girls in US special education programmes. Gender autism spectrum disorder based on DSM-5 criteria: evidence
and Education 20(4): 349–359. from clinician and teacher reporting. Journal of Abnormal
Asher SR, Rose AJ, Gabriel SW, et al. (2001) Peer rejection in Child Psychology 42(8): 1381–1393.
everyday life. In: Leary MR (ed.) Interpersonal Rejection. Hiller RM, Young RL and Weber N (2016) Sex differences in
New York: Oxford University Press, pp.105–142. pre-diagnosis concerns for children later diagnosed with
Attwood T (2006) Asperger Syndrome and Girls: The Pattern autism spectrum disorder. Autism 20(1): 75–84.
of Abilities and Development of Girls with Asperger’s Jordan-Young R (2010) Brainstorm: The Flaws in the Science of
Syndrome. Arlington, TX: Future Horizons. Sex Differences. Cambridge, MA: First Harvard University
Blatchford P, Baines E and Pellegrini A (2003) The social con- Press.
text of school playground games: sex and ethnic differences, Kasari C and Rotheram-Fuller E (2005) Current trends in psycho-
and changes over time after entry to junior school. British logical research on children with high-functioning autism
Journal of Developmental Psychology 21: 481–505. and Asperger disorder. Current Opinion in Psychiatry
Bussing R, Zima BT, Perwien AR, et al. (1998) Children in spe- 18(5): 497–501.
cial education programs: attention deficit hyperactivity dis- Kasari C, Locke J, Gulsrud A, et al. (2011) Social networks and
order, use of services, and unmet needs. American Journal friendships at school: comparing children with and with-
of Public Health 88(6): 880–886. out ASD. Journal of Autism and Developmental Disorders
Card NA, Stucky BD, Sawalani GM, et al. (2008) Direct and 41(5): 533–544.
indirect aggression during childhood and adolescence: a Kasari C, Dean M, Orlich F, et al. (2015) Children with ASD
meta-analytic review of gender differences, intercorrela- and social skills groups at school: randomized trial com-
tions, and relations to maladjustment. Child Development paring intervention approach and peer composition.
79(5): 1185–1229. Journal of Child Psychology and Psychiatry 57(2): 171–
Corsaro WA and Eder D (1990) Children’s peer cultures. Annual 179.
Review of Sociology 16: 197–220. Knickmeyer RC, Wheelwright S and Baron-Cohen S (2007)
Crick NR and Dodge KA (1996) Social information-processing Sex-typical play: masculinization/defeminization in girls
mechanisms in reactive and proactive aggression. Child with autism spectrum condition. Journal of Autism and
Development 67(3): 993–1002. Developmental Disorders 38: 1028–1035.
Cridland EK, Jones SC, Magee CA, et al. (2014) Family-focused Kopp S and Gillberg C (1992) Girls with social deficits and
autism spectrum disorder research: a review of the utility of learning problems: autism, atypical Asperger syndrome or
family system approaches. Autism 18(3): 213–222. a variant of these conditions. European Child & Adolescent
Dean M, Fox GA and Kasari C (2013) How narrative difficulties Psychiatry 1: 89–99.
build peer rejection: a discourse analysis of a girl with ASD Lai M and Baron-Cohen S (2015) Identifying the lost genera-
and her female peers. Discourse Studies 15: 147–166. tion of adults with autism spectrum conditions. The Lancet
Dean M, Kasari C, Shih W, et al. (2014) The peer relationships Psychiatry 2(11): 1013–1027.
of girls with ASD at school: comparison to boys and girls Lord C (1982) Sex differences in autism. Journal of Autism and
with and without ASD. Journal of Child Psychology and Developmental Disorders 27(5): 621–626.
Psychiatry 55(11): 1218–1225. Lord C, Risi S, Lambrecht L, et al. (2000) The Autism Diagnostic
Dhuey E and Lipscomb S (2010) Disabled or young? Relative age Observation Schedule-Generic: a standard measure of
and special education diagnosis. Economics of Education social communication deficits associated with the spectrum
Review 29: 857–872. of autism. Journal of Autism and Developmental Disorders
Dworzynski K, Ronald A, Bolton P, et al. (2012) How different are 30(3): 205–223.
girls and boys above and below the diagnostic threshold for Lord C, Rutter M, DiLavore PC, et al. (2002) Autism Diagnostic
autism spectrum disorders? Journal of the American Academy Observation Schedule—Generic. Los Angeles, CA: Western
of Child and Adolescent Psychiatry 51(8): 788–797. Psychological Services.
Dean et al. 689

Maccoby EE (1988) Gender as a social category. Developmental Pellegrini AP, Blatchford P, Kentaro K, et al. (2004) A short-
Psychology 24(6): 755–765. term longitudinal study of children’s playground games
Maccoby EE (1999) The Two Sexes: Growing up Apart Coming in primary school: implications for adjustment to school
Together. Cambridge, MA: First Harvard University Press. and social adjustment in the USA and the UK. Social
Maccoby EE (2004) Aggression in the context of gender Development 13(1): 107–123.
development. In: Putallaz M and Bierman KL (eds) Shattuck PT, Durkin M, Maenner M, et al. (2009) Timing of
Aggression, Antisocial Behavior, and Violence among identification among children with an autism spectrum dis-
Girls: A Developmental Perspective (Duke series in child order: findings from a population-based surveillance study.
development and public policy). New York: Guilford Journal of the American Academy of Child and Adolescent
Publications, pp.3–22. Psychiatry 48(5): 474–483.
McIntyre T and Tong V (1998) Where the boys are: do cross- Solomon M, Miller M, Taylor SL, et al. (2012) Autism symp-
gender misunderstanding of language use and behavior toms and internalizing psychopathology in girls and boys
patters contribute to the overrepresentation of males in pro- with autism spectrum disorders. Journal of Autism and
grams for students with emotional and behavioral disorders? Developmental Disorders 42: 48–59.
Education and Treatment of Children 21(3): 321–332. Tierney S, Burns J and Kilbey E (2016) Looking behind
Mandy W, Chilvers R, Chowdhury U, et al. (2012) Sex differ- the mask: social coping strategies of girls on the autis-
ences in autism spectrum disorder: evidence from a large tic spectrum. Research in Autism Spectrum Disorders 23:
sample of children and adolescents. Journal of Autism and 73–83.
Developmental Disorders 42(7): 1304–1313. Wilding L and Griffey S (2015) The strength-based approach
Mandy W, Roughan L and Skuse D (2014) Three dimensions to educational psychology practice: a critique from social
of oppositionality in autism spectrum disorder. Journal of constructionist and systemic perspectives. Educational
Abnormal Child Psychology 42(2): 291–300. Psychology in Practice 31(1): 43–55.