ORIGINAL PAPER
The Experiences of Late-diagnosed Women with Autism
Spectrum Conditions: An Investigation of the Female Autism
Phenotype
Sarah Bargiela
1
•
Robyn Steward
2
•
William Mandy
1
Published online: 25 July 2016
The Author(s) 2016. This article is published with open access at Springerlink.com
Abstract We used Framework Analysis to investigate the
female autism phenotype and its impact upon the under-
recognition of autism spectrum conditions (ASC) in girls
and women. Fourteen women with ASC (aged
22–30 years) diagnosed in late adolescence or adulthood
gave in-depth accounts of: ‘pretending to be normal’; of
how their gender led various professionals to miss their
ASC; and of conflicts between ASC and a traditional
feminine identity. Experiences of sexual abuse were
widespread in this sample, partially reflecting specific
vulnerabilities from being a female with undiagnosed ASC.
Training would improve teachers’ and clinicians’ recog-
nition of ASC in females, so that timely identification can
mitigate risks and promote wellbeing of girls and women
on the autism spectrum.
Keywords Autism spectrum conditions (ASC) Autism
spectrum disorder (ASD) Diagnosis Female autism
phenotype
Introduction
Autism spectrum condition (ASC), also known as ‘autism
spectrum disorder’ (ASD)
1
is a neurodevelopmental syn-
drome characterised by difficulties with social reciprocity,
social communication, flexibility and sensory processing
(American Psychiatric Association [APA] 2013). People
with ASC are at risk of a range of emotional, behavioural,
social, occupational and economic difficulties (e.g., Howlin
and Moss 2012) . The timely identification of ASC can
mitigate some of these risks and improve quality of life, for
example by identifying needs and appropriate interven-
tions, increasing access to services, making others less
judgemental of the person with ASC and their parents,
reducing self-criticism, and helping to foster a positive
sense of identity (Hurlbutt and Charmers 2002; Portway
and Johnson 2005; Ruiz Calzada et al. 2012; Russell and
Norwich 2012; Wong et al., 2015).
Compared to males, females are at substantially ele-
vated risk of their ASC going undiagnosed: their diffi-
culties are frequently mislabelled or missed entirely (Lai
and Baron-Cohen 2015). This is shown by the observa-
tion that in non-referred samples there are between two
and three males for each female with ASC (e.g., Con-
stantino et al. 2010; Kim et al. 2011; Zwaigenbaum et al.
2012); whereas in clinical samples ascertained from ASC
services, the male-to-female ratio is usually four-to-one
or higher (e.g., Fombonne 2009). Thus, many females
who, if skilfully assessed, would meet full diagnostic
criteria for ASC, never receive a diagnosis and the help
that, potentially, comes with it. Even when females with
ASC are identified, they receive their diagnosis (and
associated support) later than equivalent males (Giarelli
et al. 2010). Furthermore, compared to males, females
& William Mandy
w.mandy@ucl.ac.uk
1
Research Department of Clinical, Educational and Health
Psychology, UCL, Gower Street, London WC1E 6BT, UK
2
Centre for Research in Autism and Education, UCL Institute
of Child Health, 55-59 Gordon Square, London WC1H 0NU,
UK
1
We use the term ‘autism spectrum conditions’ (ASC) as a direct
synonym for the DSM-5 term ‘autism spectrum disorder’ (ASD). This
is in accordance with views of members of the autism community, to
be more respectful of neurodiversity. Our use of ‘ASC’ aims to
convey that people on the autism spectrum show differences that
include strengths as well as difficulties.
123
J Autism Dev Disord (2016) 46:3281–3294
DOI 10.1007/s10803-016-2872-8
require more severe autistic symptoms (Russell et al.
2010) and greater cognitive and behavioural problems
(Dworzynski et al. 2012) to meet ASC criteria, and
teachers underreport autistic traits in their female pupils
(Posserud et al. 2006). This gender bias has serious
consequences for the health and wellbeing of girls and
women with ASC, and has been identified by the autism
community as a key problem to be addressed by research
(Pellicano et al. 2014).
One proposed explanation of the ascertainment bias
against females with ASC is that there is a female autism
phenotype; a female-specific manifestation of autistic
strengths and difficulties, which fits imperfectly with cur-
rent, male-based conceptualisations of ASC. (APA 2013;
Hiller et al. 2014; Lai et al. 2015; Mandy et al. 2012).
There is an emerging evidence-base to support the exis-
tence of this female autism phenotype. For example, in line
with the reports of clinicians and people with ASC, there is
empirical evidence that girls and women with ASC show
higher social motivation and a greater capacity for tradi-
tional friendships than do males with ASC (Head et al.
2014; Sedgewick et al. 2015). Furthermore, compared to
equivalent males, females with ASC are less likely to have
externalising behaviours, such as hyperactivity/impulsivity
and conduct problems, and are more vulnerable to inter-
nalising problems, such as anxiety, depression and eating
disorders (Mandy et al. 2012; Huke et al. 2013); and
consistently score lower on measures of repetitive and
stereotyped behaviour (Van Wjingaarden-Cremers et al.
2014).
Nevertheless, research into ASC gender differences is at
an early stage, and there is currently no definitive account
of the female autism phenotype that could be used to
inform efforts to reduce the ascertainment bias against girls
and women with ASC (Lai et al. 2015). Findings on gender
differences in core diagnostic social and communication
symptoms have been inconclusive, with there being no
clear picture as to whether, compared to males with ASC,
females with ASC either show greater (e.g., Hartley and
Sikora 2009), lesser (e.g., McLennan et al. 1993) or equal
social difficulties (e.g., Mandy et al. 2012). Also, it is
uncertain whether the consistent observation that females
score lower than males for repetitive behaviours reflects
genuinely lower levels of these traits, or if female-typical
repetitive behaviours do not register on current measure-
ment tools (Van Wjingaarden-Cremers et al., 2014). Fur-
ther, it has been suggested that a key feature of the female
autism phenotype is a capacity to ‘camouflage’ social dif-
ficulties in social situations (e.g., Kenyon 2014). However,
despite promising initial investigations (e.g., Baldwin and
Costley 2015; Cridland et al. 2014; Mandy and Tchanturia
2015; Rynkiewicz et al. 2016) further work is required to
operationalize the construct of camouflaging, to prepare the
ground for the development of measures that could be used
in quantitative investigations.
These uncertainties about the nature of the female aut-
ism phenotype reflect, in part, two key methodological
challenges to doing research in this area, which have
constrained the validity of findings to date. First, most
studies have investigated participants with ASC ascer-
tained from autism clinics [e.g., see reviews by Lai et al.
(2015) and Van Wjingaarden-Cremers et al. (2014)]. This
has the effect of excluding the very participants most rel-
evant to the research, namely females who have been
missed by clinical services because their ASC exemplifies
the female autism phenotype. Second, males and females
have tended to be compared on gold-standard, well-estab-
lished measures of ASC symptoms. For most ASC research
this would be a methodological strength, but when inves-
tigating gender differences it is potentially problematic.
Such measures have been developed and validated with
largely male samples, and may lack sensitivity to the
female autism phenotype. Therefore, there is a need for
measures that are demonstrably sensitive to autistic fea-
tures as they present in females as well as males (Lai et al.
2015). Both of the methodological problems described
above would have the effect of underestimating ASC
gender differences. A further consideration is that no
research to date has sought to examine directly how the
female autism phenotype can lead to a situation whereby a
girl or woman meets criteria for ASC, but is missed by
professionals. Such investigations will be informative for
those seeking to improve diagnostic practice to reduce
gender-based inequities in ASC care.
Given the above, in order to advance the study of ASC
gender differences we conducted a study with three key
features. First, we aimed to investigate directly not only the
nature of the female autism phenotype, but also how it
impacts upon risk of a girl and/or woman’s ASC going
unrecognised. Second, we recruited women with ASC
whose autistic difficulties had gone unrecognised in
infancy, childhood and early adolescence. We reasoned
that such late-diagnosed individuals would be more likely
to exemplify elements of the female autism phenotype that
are under-represented in samples of those identified in a
timely fashion, and can provide insights into how such
characteristics led to them being missed by clinical ser-
vices. This approach is supported by the recent finding that
the gender ratio in adult ASC clinics is lower (two males to
one female) than in child and adolescent services (five
males to one female), suggesting that later-diagnosed
samples are most likely to include a representative sample
of females with ASC (Rutherford et al. 2016). Third, we
took an inductive (i.e., data-driven) approach, conducting a
qualitative investigation. Our aim was not to test
hypotheses about the female autism phenotype by formally
3282 J Autism Dev Disord (2016) 46:3281–3294
123
comparing males and females. Rather, we sought to gen-
erate new ideas and deepen understanding of key concepts,
such as ‘camouflaging’ (Barker and Pistrang 2015). This
work is designed to yield novel, well-defined hypotheses
about the female autism phenotype to guide future quan-
titative investigations; and to promote the development of
measures that capture female as well as male manifesta-
tions of ASC.
Framework Analysis (Ritchie et al. 2003), widely
applied in health research to generate theory and promote
the development of new measures, was used to analyse
qualitative data from in-depth interviews in order to
address the following questions:
1. What is the nature of the female autism phenotype, as
experienced by late-diagnosed women with ASC?
2. How does the female autism phenotype influence
young women’s experiences of diagnosis, misdiagno-
sis and missed diagnosis?
3. How do late-diagnosed women with ASC adapt in
response to the challenges they face?
Methods
Participants
Participants were 14 women with ASC. Women were eli-
gible to participate in the study if they met the following
inclusion criteria: (1) aged between 18 and 35 years; (2)
diagnosed with ASC by a certified professional (psychia-
trist, clinical psychologist) in the UK National Health
Service; (3) ASC diagnosis was received in late adoles-
cence or adulthood (aged 15 years or older); (4) ASC
diagnosis had been received within 10 years of study par-
ticipation; (5) living in the United Kingdom; (6) without an
intellectual disability, as indicated by having an IQ above
70. Age and IQ limits for the sample were set to limit group
heterogeneity, as these variables likely condition the
experiences of women with ASC. Seventeen women con-
tacted the researcher asking to participate in the study.
Three did not meet eligibility criteria: one was outside of
the qualifying age range (51 years) and two lived outside of
the UK (USA and Australia).
Participant characteristics are shown in Table 1. At the
time of the study, seven of the participants reported being
able to support themselves through paid employment. One
had a profession, but was signed off work due to mental
health problems. The three full-time students were sup-
ported by research grants or family. The remaining par-
ticipants did volunteer work or full-time parenting. The
mean age of the participants was 26.7 years, (SD = 2.3).
All their ASC diagnoses had been made since 2004, and
the mean age of diagnosis was 21.3 years (SD = 4.8).
Thirteen had received their diagnosis from a specialist
autism service, having never previously been assessed for
ASC. One participant was diagnosed after assessment in a
Child and Adolescent Mental Health Service, also without
any previous history of autism assessment. The research
team did not retrieve clinical records to double-check the
veracity of self-reported diagnoses, but every participant
scored above cut-off on the AQ-10. All participants had
estimated IQs above 70 on the Wechsler Test of Adult
Reading.
Table 1 Characteristics of the sample
Participant Age at time of interview Age at diagnosis IQ AQ-10 (cut-off = 6) Employment
P01 23–26 years 19–22 years 122 9 Illustrator
P02 23–26 years 15–18 years 115 6 Full time Student
P03 27–30 years 27–30 years 110 8 Health Professional
P04 22–26 years 15–18 years 113 10 Support Worker
P05 22–26 years 22–25 years 122 10 Volunteer
P06 27–30 years 23–26 years 124 9 Full time Student
P07 19–22 years 19–22 years 124 9 Fine Artist
P08 22–26 years 19–22 years 85 10 Health Professional
P09 27–30 years 19–22 years 108 8 Administrator
P10 27–30 years 23–26 years 117 9 Volunteer
P11 27–30 years 27–30 years 108 8 Full-time mother
P12 27–30 years 19–22 years 115 8 Full-time Student
P13 23–26 years 19–22 years 92 10 Volunteer
P14 27–30 years 23–26 years 110 9 Athlete
Precise age and age at time of diagnosis are not given to protect participant confidentiality
AQ-10 = 10 item version of the Autism Spectrum Quotient
J Autism Dev Disord (2016) 46:3281–3294 3283
123
Procedure
Participants were recruited via existing contacts within the
research team and through adverts placed on social media
sites frequented by women with ASC. Information sheets
were provided to eligible participants at least 24 hours
before consent was sought, and only those who subse-
quently gave full informed consent joined the study. All 14
participants completed quantitative measures of IQ,
depression, anxiety and autistic traits (see Measures) in
order to ‘situate the sample’, i.e. to give the reader a sense
of who the participants where, to inform thinking about
transferability of the findings (Barker and Pistrang 2015).
Nine interviews were conducted face-to-face, four partici-
pants were interviewed via Internet videoconferencing, and
one interview was completed on the telephone. Video-
conference and telephone interviews were conducted in
order to avoid excluding participants who did not want a
face-to-face interviews, due to anxiety, sensory issues and/
or a reluctance to engage in direct social interaction.
Measures
Semi-structured Interview
Data were collected using a semi-structured interview
schedule developed specifically for this study by the
research team, via a process of consultation with women
with ASC, expert clinicians and researchers. Thus, the
topics covered in the interview reflected the research aims,
previous research, clinical insights into the female autism
phenotype and the priorities of members of the autism
community. In line with guidelines for semi-structured
interviews, it was designed to be used flexibly, thus
allowing the interviewer freely to follow the participant’s
line of response, maximising the chances of collecting
valid, in-depth data from people with atypical social
communication (Smith 1995). The interview began by
asking women to tell the story of how they were diagnosed
with ASC, including exploration of their sense of how their
gender had impacted upon this. It then moved to consid-
eration of interests, social relationships, sensory experi-
ences and mental health. The final questions concerned the
participants’ perceptions of ASC gender differences. A
copy of the interview schedule is available by request from
the first author (SB).
Autism Quotient-10 (AQ-10)
This ten-item, self-report, brief version of the Autism
Spectrum Quotient was used to confirm clinical diagnosis,
and to give an approximate measure of ASC severity
(Allison et al. 2012).
General Health Questionnaire-12 (GHQ-12)
The GHQ-12 is a twelve-item, brief version of the GHQ-
60. It is used as a screening device to assess the respon-
dent’s current mental state (Goldberg and Williams 1988)
and is a reliable and valid measure of severity of psycho-
logical difficulties (Goldberg et al. 1997). The GHQ-12 has
a maximum score of 36. For young adults, 24 has been
established as an optimal cut-point for identifying mental
health difficulties (Makowska et al. 2002). Participants
were asked to complete the measure for the researchers to
gain an overview of the mental wellbeing of the sample.
Hospital Anxiety and Depression Scale (HADS)
The HADS is a fourteen-item self-assessment scale used to
detect depression (seven items) and anxiety (seven items)
over a 1-week period (Snaith 2003). Participants are asked
to respond to statements that best describe their state of
mind, with higher scores indicating greater symptom
severity. Both the depression and anxiety subscales have a
maximum score of 21, with scores of 8 and above being
indicative of clinically severe difficulties.
Wechsler Test for Adult Reading (WTAR)
The WTAR is a neuropsychological assessment tool, which
provides a reliable, valid and low intensity estimate of
intelligence (Holdnack 2001). In this study, researchers
used the WTAR in order to offer a quick estimate of verbal
IQ, to establish whether participants were able to easily
verbalise their experiences, to situate the sample and to
check that participants met the inclusion criterion of not
having an intellectual disability.
Data Analysis
All interviews were transcribed verbatim, and Framework
Analysis was applied to the data (Ritchie and Spencer 1994;
Ritchie et al. 2003). This method involves following a
structured sequence of steps in order to systematically
identify themes within qualitative data. Framework Analysis
was chosen as it is a widely used and transparent method of
qualitative analysis that allows researchers to generate new
theory from data whilst focusing their inquiries on pre-de-
termined research objectives. We adhered to guidelines for
good practice in qualitative research (Mays and Pope 2000;
Pope et al. 2000; Stiles 1999) by conducting the following
credibility checks, to ensure that interpretations of the data
were sound and fair. First, to avoid relying upon a single
researcher’s perspective, a consensus approach (Barker and
Pistrang 2005) was employed: SB took the lead in analysis
but all three authors regularly discussed themes until,
3284 J Autism Dev Disord (2016) 46:3281–3294
123
eventually, they derived a final set of themes and subthemes.
Second, during this process, WM audited the framework
against transcript data (Elliott et al. 1999). Finally, respon-
dent credibility checks were conducted to promote testi-
monial validity (Barker and Pistrang 2005), whereby the
framework was sent to participants for their feedback, to
ensure it reflected their experiences.
Results
Anxiety, Depression and Wellbeing in the Sample
All fourteen participants completed the HADS and GHQ-12.
The mean HADS-A (anxiety) score was 13.5 (SD 3.7, range
7–20), which lies above the recommended clinical cut-off.
All but one participant scored above the cut-off score. The
mean HADS-D (depression) score was 5.3 (SD 4.8, range
1–15), below the clinical threshold, with three participants
scoring in the clinical range. The mean GHQ-12 score was
15.4 (SD 4.4, range 9–27), below the cut-off indicating
mental disorder. Three participants scored in the ‘distress’
range and a further two fell in the ‘severe’ range, indicative
of severe psychological difficulties at the time of interview.
Qualitative Analysis: Themes and Sub-themes
Relevant data in the transcripts were organised into themes
and subthemes, with the final thematic framework
presented in Table 2. Four themes, comprising nineteen
subthemes, were identified. Quotes were labelled as ‘P’
followed by a unique identifier for each participant
(Table 1). The four major identified themes are as follows:
(1) ‘You’re not autistic’, which recognises common barri-
ers to gaining a diagnosis as a woman; (2) ‘Pretending to be
normal’ which identifies strategies that young women
employed when trying to fit in with their peers; (3) ‘Passive
to assertive’ explores how passivity and social naivety
impact on young women with ASD and how they have
learnt to be assertive; and (4) ‘Forging an identity as a
young woman with ASC’ outlines common difficulties
associated with being female with a social communication
disorder and the protective role played by special interests.
‘‘You’re not autistic’’
This theme included reported experiences of autistic diffi-
culties being ignored and misunderstood, perceived reasons
for this, and beliefs about the implications of having
received a late diagnosis. Almost all the young women
reported having experienced one or more mental health
difficulty, with anxiety, depression and eating disorder
being the most commonly reported. Most participants
commented that health professionals treating them had not
noticed their symptoms might be related to ASC:
‘‘Four to five years of depression and anxiety treat-
ment…years of talking therapy…and not once did
Table 2 - Framework analysis, including frequency of themes
Theme Subtheme Frequency
‘‘You’re not autistic’’ Labelled with non-autistic diagnoses 12
Unhelpful professional stereotypes of autism 12
Quiet at school, so went unnoticed: ‘I should have burnt more cars’ 6
Misunderstood, unsupported or blamed by teachers 8
The costs of a late diagnosis 8
Pretending to be ‘normal’ ‘Wearing a mask’ 8
Learning social behaviours from TV, books and magazines 6
Social mimicry 5
The costs of masking 5
Passive to assertive Please, appease, avoid conflict 7
Entrapment in abusive relationships or risky situations 8
Victim of sexual abuse 9
Learning to be assertive 8
Forging an identity as a woman with ASD Societal pressures: what is expected of young women 11
Friendships: uncertainty and intensity 12
Hard to be friends with neurotypical girls 8
Easier to be friends with boys 7
Friendships and support online 7
Interests define identity and self confidence 7
J Autism Dev Disord (2016) 46:3281–3294 3285
123