Journal of Autism and Developmental Disorders

https://doi.org/10.1007/s10803-018-3792-6

ORIGINAL PAPER

Development and Validation of the Camouflaging Autistic Traits

Questionnaire (CAT-Q)
Laura Hull1,6   · William Mandy1 · Meng‑Chuan Lai2,3,4 · Simon Baron‑Cohen3 · Carrie Allison3 · Paula Smith3 ·
K. V. Petrides5

© The Author(s) 2018

Abstract
There currently exist no self-report measures of social camouflaging behaviours (strategies used to compensate for or mask
autistic characteristics during social interactions). The Camouflaging Autistic Traits Questionnaire (CAT-Q) was developed
from autistic adults’ experiences of camouflaging, and was administered online to 354 autistic and 478 non-autistic adults.
Exploratory factor analysis suggested three factors, comprising of 25 items in total. Good model fit was demonstrated through
confirmatory factor analysis, with measurement invariance analyses demonstrating equivalent factor structures across gender
and diagnostic group. Internal consistency (α = 0.94) and preliminary test–retest reliability (r = 0.77) were acceptable. Con-
vergent validity was demonstrated through comparison with measures of autistic traits, wellbeing, anxiety, and depression.
The present study provides robust psychometric support for the CAT-Q.

Keywords  Autism · Camouflaging · Masking · Compensation · Coping Strategies

Social camouflaging is defined as the use of strategies by behaviours when assessing autism in the newly released
autistic people to minimise the visibility of their autism 11th edition of the International Classification of Diseases
during social situations (Lai et al. 2011). This topic has (Zeldovich 2017), but has been described by autistic people
recently come to the attention of researchers, as recognised and clinicians for many years. It may be a widespread and
by the call for clinicians to be aware of masking or coping important phenomenon in autism, especially in intellectu-
ally able individuals. We note here that, following prefer-
ences from a majority of the autism community (Kenny
Electronic supplementary material  The online version of this
article (https​://doi.org/10.1007/s1080​3-018-3792-6) contains et al. 2015), we use identity-first language in this paper (e.g.
supplementary material, which is available to authorized users. ‘autistic person’) while recognising that some individuals
prefer the use of person-first language (e.g. ‘person with
* Laura Hull autism’). Social camouflaging encompasses an explicit effort
laura.hull.14@ucl.ac.uk
to ‘mask’ or ‘compensate’ for autistic characteristics; and
1
Research Department of Clinical, Educational & Health to use conscious or unconscious techniques which result in
Psychology, University College London, London, UK a less autistic behavioural presentation (Hull et al. 2017;
2
Child and Youth Mental Health Collaborative, Centre Lai et al. 2017; Livingston and Happé 2017). Examples of
for Addiction and Mental Health and The Hospital for Sick camouflaging behaviours described in the current literature
Children, Department of Psychiatry, University of Toronto, include forcing oneself to make eye contact during a social
Toronto, ON, Canada
interaction, or pretending that one is doing so by looking
3
Autism Research Centre, Department of Psychiatry, at the space between someone’s eyes or at the tip of their
University of Cambridge, Cambridge, UK
nose, or using working memory strategies to develop a list
4
Department of Psychiatry, National Taiwan University of appropriate topics for conversation. Camouflaging is
Hospital and College of Medicine, Taipei, Taiwan
driven by the desire to ‘fit in’ so as to appear non-autistic,
5
London Psychometric Laboratory, University College and to form relationships with others, which may be harder
London, London, UK
to achieve when the person presents autistic behaviour (Hull
6
Department of Psychology, University College London, 26 et al. 2017).
Bedford Way, London WC1H 0AP, UK

13
Vol.:(0123456789)
 Journal of Autism and Developmental Disorders

The concept of social camouflaging was first investigated and their overt behaviours (‘external autistic presentation’).
through qualitative research with autistic girls and women, This approach was used by Lai et al. (2017), who measured
looking in particular at reasons why these individuals may the discrepancy between self-reported autistic traits on the
not be diagnosed until later in life. Themes identified in this Autism Spectrum Quotient (AQ) as well as social cognitive
research include the concept of ‘masking’, or portraying abilities on the Reading the Mind in the Eyes test (RMET)
a non-autistic persona (Bargiela et al. 2016; Tierney et al. (internal autistic status) and scores on the Autism Diagnostic
2016), and the idea that through copying others and control- Observation Schedule (ADOS) (external autistic presenta-
ling behavioural expression, autistic girls and women could tion). The study found a greater AQ/RMET-ADOS discrep-
compensate for some of the social and communication dif- ancy score for female than male participants, and that higher
ficulties they experienced (Dean et al. 2017). This qualitative discrepancy scores were associated with greater depressive
research suggested that there may be some negative conse- symptoms for men, but not for women.
quences of camouflaging. These include links to heightened A similar result was found by Ratto et al. (2017), where
stress or mental health conditions such as depression, and autistic females with higher IQ were less likely to meet diag-
reduced access to clinical support and services as a result of nostic criteria on the Autism Diagnostic Interview (Revised)
difficulties being hidden (Cage et al. 2017; Head et al. 2014). than males, despite being matched on ADOS scores and
Autistic females’ camouflaging may even account for the having higher levels of parent-report functioning difficul-
later and less frequent diagnoses of females than males with ties than males. Other studies have demonstrated autistic
the same autistic characteristics (Begeer et al. 2013; Giarelli females’ greater use of camouflaging strategies during com-
et al. 2010; Rutherford et al. 2016). In addition, some quali- munication than males, whether through gesture (Rynkie-
tative research that has begun to look at the experiences of wicz et al. 2016), or filling pauses in conversation (Par-
camouflaging amongst autistic men suggests that while both ish-Morris et al. 2017), despite overall comparable social
men and women may camouflage their autism, there might skills. These methods measure camouflaging by identifying
be gendered differences in both the techniques used and the discrepancies between different measures of social ability
consequences of camouflaging (Hull et al. 2017). or autistic characteristics, such that individuals (especially
This qualitative research has offered new insights into females) appear less autistic in some settings yet still meet
under-investigated social behaviours in autism, and has autism diagnostic criteria in others.
raised important questions to address: Who, among the A strength of these discrepancy approaches is their con-
many different autistic people, camouflages their autism? ceptual rigour, as they seek to operationalize the key fea-
Do autistic girls and women camouflage more than boys and ture of camouflaging; that it is a dissociation between an
men, and does this partly account for gender disparities in individual’s experience of being autistic and the behaviours
the rate and timing of diagnosis (Begeer et al. 2013; Loomes they portray to the outside world. In addition, using autism
et al. 2017)? What is the relationship between camouflaging assessment tools as a measure of external autism presenta-
and mental health outcomes? Quantitative investigation of tion demonstrates the impact camouflaging can have in a
these questions has to date been hindered by the challenges clinical setting, especially for autistic women. However, a
of accurately measuring camouflaging. key disadvantage is that this approach relies upon an index
of how autistic a person is (i.e., their internal autistic status)
independent of their behavioural presentation. Given that
Measures of Social Camouflaging autism is currently conceptualised at the behavioural level
because there are no reliable biomarkers for the condition
In recent years there have been some attempts to quantify (American Psychiatric Association 2013; Loth et al. 2015),
social camouflaging by autistic people. The resultant instru- this represents a significant conceptual and practical chal-
ments reflect different ways of defining and operationalising lenge. Performance on tests of cognition relevant to autism,
camouflaging, leading to some overlap but also some dis- or scores on self-reported measures of autism traits can only
crepancies in how camouflaging behaviours are measured. ever be a proxy measure of internal autistic status: we cur-
Livingston and Happé (2017) suggest that camouflaging rently have no way to identify how autistic an individual is
is a component of compensation, the “processes contributing meaningfully and accurately.
to improved behavioural presentation of a neurodevelopmen- In addition, the measurement of camouflaging using dis-
tal disorder, despite persisting core deficit(s) at cognitive crepancy approaches does not allow for unsuccessful camou-
and/or neurobiological levels” (p. 8), and hence should be flaging attempts to be assessed. Some autistic people may use
measured at the behavioural, cognitive, and neurobiological a variety of strategies in an attempt to appear less autistic to
levels. We call these approaches to measuring camouflag- others, but these may be only partially or not at all success-
ing ‘discrepancy methods’, as they seek to measure the gap ful. This is especially important when considering the evi-
between how autistic a person is (‘internal autistic status’) dence for a link between self-reported camouflaging and poor

13
Journal of Autism and Developmental Disorders

mental health (Cage et al. 2017; Hull et al. 2017). If individu- camouflaging behaviours. Here, camouflaging is conceptu-
als attempt to camouflage but are ultimately unsuccessful this alised based on the reported experiences of individuals who
may further increase the social, emotional, and psychological have (and have not) camouflaged their autism, and the behav-
harm resulting from their camouflaging efforts. iours and intentions described by these individuals are used to
An alternative to the discrepancy approaches described develop a list of camouflaging strategies to measure. Autistic
above is one based on observational recognition of camou- individuals can then report directly on their own camouflaging
flaging; measuring the specific behaviours and experiences behaviours, identifying strategies and intentions that might not
which represent camouflaging. Such ‘observational/reflective be visible to an observer without in-depth discussion with the
methods’ circumvent the limitation of being unable to meas- autistic person themselves.
ure an individual’s internal autistic state. Camouflaging can This self-report method, based on an observational/reflec-
be measured consistently and compared between individuals, tive approach, has several strengths. First, identifying cam-
and behaviours can be identified regardless of how successful ouflaging behaviours based on strategies reported by autis-
they are. In other words, identification of camouflaging is not tic individuals reduces the potential for introducing bias via
reliant on either a proxy measure of internal autistic status, or researchers’ and clinicians’ perceptions of autistic behaviours
the need to display a typical social presentation. and abilities. Autistic adults have previously reported being
Dean et al. (2017) used an observational/reflective approach told by clinicians that their ability to camouflage (for exam-
to identify camouflaging strategies used by autistic girls when ple, by making or appearing to make eye contact) meant they
interacting with peers through behavioural observation. could not be autistic, despite meeting autism diagnostic criteria
Behavioural techniques, such as standing near to peers who in other ways (Hull et al. 2017). Clinicians and researchers
are interacting, but not actually engaging in with them, were may only observe autistic individuals in one structured and
classified as camouflaging strategies, and were observed more limited situation and so may not identify certain behaviours
in autistic girls than in autistic boys or non-autistic girls, in a as camouflaging strategies, whereas autistic individuals and
school setting. This led to the superficial appearance of suc- those who know them well have a unique insight into their own
cessful social interaction, but did not actually result in friend- behaviours across a variety of situations. Second, self-report
ships or sustained engagement for the autistic girls using these measures of camouflaging allow for operationalisation of the
techniques. Dean et al. operationalisation of camouflaging is attempt to camouflage—the intention put into camouflaging
based on the idea of blending into the social environment, a autistic characteristics, and the techniques used, which may
strength of which is that the need to camouflage may vary not result in any observable external change for someone who
depending on the situation. In addition, this definition of cam- does not know the person well.
ouflaging emphasises that camouflaging behaviours may be Both the discrepancy and observational/reflective
learned or mimicked from non-autistic peers. approaches described above offer ways to define and there-
This approach to camouflaging has the advantage of fore measure camouflaging in autism. All the methods used or
allowing for variation in camouflaging behaviours and their suggested have their own strengths and weaknesses, thus com-
success. Techniques learned and used in some situations bining multiple methods in a triangulation approach allows for
may not be successful in others, and an individual’s overall greater accuracy in measuring and identifying a complex phe-
camouflaging ability may partly depend on their ability to nomenon such as camouflaging (Thurmond 2001). Participant
adapt to different situations. The cognitive flexibility ena- report is needed to identify intention to camouflage, behav-
bling this has already been suggested as one explanation for ioural observation to identify how successful that camouflag-
autistic girls’ superficially higher social skills (Lehnhardt ing is, and measures of cognitive traits and autistic character-
et al. 2015). However, this measure of camouflaging is based istics to identify how much the person is camouflaging their
on non-autistic observers’ ideas of what camouflaging looks underlying ‘autistic-ness’ and how they do or do not achieve
like. Intentions and behaviours of camouflaging which clini- this. Methods for measuring behavioural camouflaging, and
cians and researchers may not be aware of, but which may cognitive and autistic-like traits, already exist or have been
form an important part of autistic individuals’ camouflaging proposed (Dean et al. 2017; Lai et al. 2017; Livingstone and
strategies, have not yet been measured. Happé 2017); however until now, no self-report measures of
camouflaging behaviours have been developed.

Self‑Reported Measurement
of Camouflaging Camouflaging Across the Dimensions

Another observational/reflective approach to the operation- Autism is a dimensional characteristic; traits are distributed
alisation of camouflaging addresses some of these remain- across the entire population, but with a cut-off point at the
ing issues: asking autistic people themselves about their extreme end requiring clinical identification and support

13
 Journal of Autism and Developmental Disorders

(Constantino 2011; Ruzich et al. 2015; Skuse et al. 2005). Individuals with more autistic-like traits are likely to cam-
All individuals in the general population have some level of ouflage those traits to a greater extent, although this has
autistic traits, and those with an above average number may not been tested empirically before. Camouflaging has also
also camouflage these to varying extent. Camouflaging is been associated with increased social anxiety and general
similar to impression management, where behaviours which anxiety, and decreased wellbeing, in qualitative reports
occur in front of others are manipulated in order to make (Hull et al. 2017), as well as with increased depression in
a better impression (Leary and Kowalski 1990). Autistic quantitative research (Cage et al. 2017; Lai et al. 2017).
individuals engage in impression management to a lesser Accordingly, convergent validity was explored by test-
degree than non-autistic individuals (Cage et al. 2013). The ing the correlation between camouflaging and autistic-
combination of underlying autistic characteristics and extent like traits, social anxiety, general anxiety, wellbeing, and
of (successful) camouflaging produces an external ‘autistic’ depression.
presentation, with corresponding variation in general func-
tioning (Livingston and Happé 2017). Thus, it is important
to develop measures of camouflaging that are appropriate
for both autistic and non-autistic populations. Methods

Participants
The Present Study
Validation of the CAT-Q was conducted in autistic and
A psychometrically sound self-report measure of camouflag- non-autistic samples which were recruited separately.
ing behaviours is needed to improve current understanding Autistic participants were recruited via social media,
of the nature, causes and consequences of social camouflag- through the Cambridge Autism Research Database
ing. Furthermore, existing methods of measuring camouflag- (CARD), and through word-of-mouth. Non-autistic partic-
ing behaviours have not been validated in both autistic and ipants were recruited via social media and through word-
non-autistic populations. of-mouth. Participants who self-reported as autistic were
The aim of this study is therefore to develop, psycho- asked to detail the type of diagnosis, (e.g. Autism, Asper-
metrically evaluate, and validate a self-report measure of ger’s Syndrome, Autism Spectrum Disorder), the age they
social camouflaging behaviours (henceforth referred to as were diagnosed, and the type of healthcare professional
the Camouflaging Autistic Traits Questionnaire; CAT-Q), who diagnosed them. Those who reported being self-diag-
appropriate for both autistic and non-autistic populations. nosed were automatically excluded from the study and did
not complete any further questions. All participants were
Development at or above the legal age to give informed consent on their
own behalf in the UK (16 years).
Preliminary items for the CAT-Q were developed from Of those autistic participants who reported the age
qualitative responses to a previous study, and were added to they were diagnosed, 12% were diagnosed in childhood
and refined by all the authors and several external experts. (0–17 years) and 72% were diagnosed in adulthood (18 years
and over). Of those diagnosed in childhood, 38% were diag-
Psychometric Evaluation nosed by a psychiatrist, 25% by a clinical psychologist, 8%
by other specialists including neurologists and specialist
Exploratory and confirmatory factor analyses were used to nurses, 5% by a multi-disciplinary team, 2% by a Speech
identify, refine, and test the underlying factor structure of the & Language Therapist, 2% by their school, and 2% by a
CAT-Q in two separate samples. Multi-group measurement paediatrician. Of those diagnosed in adulthood, 55% were
invariance analyses were used to compare the underlying diagnosed by a clinical psychologist, 35% by a psychiatrist,
factor structure in the male and female autistic and non- 3% by a multi-disciplinary team, 3% by other specialists,
autistic samples. 0.7% by a Speech and Language Therapist, 0.7% by a GP,
Internal consistency of the measure was estimated using and 0.3% by an occupational therapist.
Cronbach’s alpha, and test–retest reliability was established In the autistic sample, 14% were aged 16–25, 23% were
by re-sending the CAT-Q to a subsample of 30 autistic par- aged 26–35, 20% were aged 36–45, 13% were aged 56–65,
ticipants approximately 3 months after they first completed 3% were aged 66–75, and 0.3% were aged 75 or over. In the
the survey. non-autistic sample, 59% were aged 16–25, 16% were aged
Convergent validity of the new measure was determined 26–35, 8% were aged 36–45, 9% were aged 46–55, 6% were
by comparing camouflaging scores with scores on theo- aged 56–65, 1% were aged 66–75, and 0.2% were aged 75
retically related constructs (Cronbach and Meehl 1955). or over (proportions may not add up to 1 due to rounding).

13
Journal of Autism and Developmental Disorders

Measures BAPQ as a measure of autism-related characteristics, rather

than as a screening tool for autism. Mean BAPQ scores were
Camouflaging Autistic Traits Questionnaire (CAT‑Q) compared for autistic and non-autistic samples and were
found to be significantly different (t[743] = 21.23, p < .001,
The measure’s operationalisation of social camouflaging is d = 1.56), with means of 4.31 (SD = 0.69) for autistic par-
based on the analysis and theoretical model described in the ticipants and 3.18 (SD = 0.73) for non-autistic participants.
qualitative study by Hull et al. (2017). Items for the CAT-Q This suggests that, although the BAPQ was designed for
were identified through multiple routes. A previous study relatives of those with autism, there were no ceiling effects
which asked for autistic adults’ experiences of camouflaging in the autistic sample.
(Hull et al. 2017) also asked participants to describe specific
behaviours they used while camouflaging. These responses Social Anxiety Scale (LSAS; Liebowitz 1987)
were refined to produce a list of behaviours reflecting the
two core components of camouflaging identified previously: A 24-item self-report questionnaire measuring social anxiety
compensation (i.e. finding ways around the social and com- in the general population. The scale requires participants to
munication difficulties associated with autism), and masking imagine being in different social situations (such as talk-
(i.e. hiding aspects of one’s autistic presentation, or present- ing to a sales assistant in a shop) and asks how much fear
ing a non-autistic persona to others). Additional camouflag- they would experience and how much they would avoid the
ing behaviours were suggested by autism experts, including situation. The LSAS has demonstrated good test–retest reli-
researchers, clinicians, and autistic adults who were con- ability and discriminant validity (Baker et al. 2002). A total
sulted directly. of 708 participants (284 adults with autism and 424 adults
Once the behaviours were identified, items that described without autism) completed the LSAS. In the total sample of
them, including reverse-coded items describing the opposite this study, internal consistency was high (α = 0.97).
of these behaviours, were developed. Items were removed
or added to ensure there was a roughly even number tapping Warwick–Edinburgh Mental Wellbeing Scale (WEMWBS;
into ‘compensation’ and ‘masking’. A total of 48 items were Tennant et al. 2007)
produced for inclusion in the study. Participants responded
using a seven-point Likert scale, from ‘Strongly Disagree’ to A 14-item self-report questionnaire measuring general well-
‘Strongly Agree’ with each statement.A total of 832 partici- being in the last 2 weeks. The WEMWBS has demonstrated
pants (354 adults with autism and 478 adults without autism) acceptable validity and reliability (Trousselard et al. 2016).
completed the CAT-Q. A total of 713 participants (289 adults with autism and 424
adults without autism) completed the WEMWBS. Internal
Broad Autism Phenotype Questionnaire (BAPQ; Hurley consistency in the total sample was high (α = 0.92).
et al. 2007)
Patient Health Questionnaire (PHQ‑9; Kroenke et al. 2001)
A 36-item self-report measure of traits associated with the
broader autism phenotype (BAP). BAP characteristics are A 9-item self-report questionnaire of depressive symptoms
associated with greater genetic liability for autism, and are in the last 2 weeks, with a clinical cut-off point of 10 for
found across the population and at especially high levels moderate depression. The PHQ-9 has demonstrated good
in relatives of those with an autism diagnosis. Scores for sensitivity and specificity for depressive symptoms (Kroenke
the total questionnaire and three sub-factors (Aloofness, et al. 2001). The PHQ-9 was only administered to autistic
Pragmatic Language, and Rigidity) are averaged across the individuals. A total of 290 autistic participants completed
36 items in the total questionnaire and 12 in each factor, the PHQ-9. Internal consistency in the autistic sample was
to produce values in a range of 0–6. A total of 744 par- acceptable (α = 0.89).
ticipants (299 autistic and 445 non-autistic) completed the
BAPQ. The BAPQ has good sensitivity (Sasson et al. 2013) Generalised Anxiety Disorder (GAD‑7; Spitzer et al. 2006)
and specificity (Hurley et al. 2007). Internal consistency
of the BAPQ in the current study (total sample) was high A 7-item self-report measure of generalised anxiety symp-
(α = 0.96). Although the BAPQ was initially developed as toms in the last 2 weeks. The GAD-7 has a clinical cut-
a measure of autistic-like traits in relatives of those with off point of 10 points and demonstrates good sensitivity
autism, it has also been used to measure autistic-like traits in and specificity (Spitzer et al. 2006). The GAD-7 was only
autistic and non-autistic groups (Ingersoll et al. 2011; Nishi- administered to autistic individuals. A total of 289 autistic
yama et al. 2014; although see Piven and Sasson 2014 for participants completed the GAD-7. Internal consistency in
an evaluation of this approach). In this case we included the the autistic sample was high (α = 0.92).

13
 Journal of Autism and Developmental Disorders

Procedure values of 0.08 or lower indicate acceptable fit (Hu and

Bentler 1999).
Participants followed a link to the online survey, hosted by
Qualtrics, where they read the information sheet and, after
contacting the researchers to answer any questions, com- Multi‑group Measurement Invariance
pleted a consent form. They then completed demographic
questions and questionnaires. The total sample was recombined and multi-group meas-
Participants who had given contact details to research- urement invariance analysis used to determine whether the
ers were contacted again 3 months later to ask them to same latent variables were measured across four groups:
re-take the questionnaire for the purpose of estimating male autistic, female autistic, male non-autistic, and
test–retest reliability. At that time, adverts were also female non-autistic. Participants who identified as a non-
placed on social media inviting autistic participants who binary gender or did not report their gender were excluded
had previously completed the survey to complete it again. from this analysis (n = 92).
Tests of measurement invariance involve the compari-
son of multiple, nested models (Sass 2011) measuring:
Analyses
(1) Configural Invariance (whether factor structure is
equal across groups); (2) Metric Invariance (whether item
All analyses were performed in R (R Core Team 2013).
loading on each factor is equal across groups); (3) Scalar
The total sample was split in two, with the first half
Invariance (item intercepts are equal across groups); and
utilised for exploratory factor analysis to identify an initial
(4) Residual Invariance (item residuals are equal across
factor structure from which a 25-item final scale was pro-
groups). Each model is compared to the previous in a for-
duced (‘exploratory sample’; N = 402), and the remainder
ward approach to first establish invariance across groups,
utilised for confirmatory factor analysis (‘confirmatory
and then test whether non-invariance has been identified
sample’; N = 430). These two samples had comparable lev-
at each additional level. ∆CFI of less than 0.01 is gen-
els of autistic-like traits; however the confirmatory sample
erally used as the most reliable marker of invariance, as
was significantly younger on average (partial η 2 = 0.13),
Χ2 values can be influenced by sample size (Cheung and
and contained proportionally more males (Cramer’s
Rensvold 2002). Diagonally Weighted Least Square Means
V = 0.12), than the exploratory sample.
(WLSM) estimators were again used, and robust statistics
are reported for all results. Satorra-Bentler scaled correc-
Exploratory Factor Analysis tions for multiple comparisons were used.

Principle components analyses using oblique rotation

were performed on the total exploratory sample (N = 402), Reliability and Validity
and separately in the autistic (N = 200) and non-autistic
(N = 202) subsamples. Retention of items was based on Internal consistency, test–retest reliability, and convergent
combined evaluation of the scree plot, following Cattell validity of the final scale were assessed in a subset of the
(1966); eigenvalues over 1.0; and parallel analysis tech- total sample who had also provided complete responses to
niques to model factor structure (Hayton et  al. 2004). at least one of the other measures in the study (N = 706;
Items with loadings below 0.40, or with cross-loadings of Autistic N = 306, Non-Autistic N = 400). Internal consist-
greater than 0.40 were excluded. ency was measured using Cronbach’s alpha, and test–retest
reliability using Pearson’s r and intra-class coefficients
(ICCs). Two-way consistency ICC was used to evaluate
Confirmatory Factor Analysis
absolute consistency between the first and second comple-
tion of the questionnaire, (McGraw and Wong 1996), with
Diagonally Weighted Least Square Means (WLSM) esti-
unity reflecting complete consistency on all items between
mators were used to take into account the ordinal nature of
time one and time two. Values of 0.50 to 0.75 indicate
the Likert-based responses (DiStefano and Morgan 2014;
moderate reliability, while values of 0.75 and above indi-
Wang and Cunningham 2005). The key indices used to
cate good reliability (Koo and Li 2016).
assess goodness-of-fit were Comparative Fit Index (CFI),
Convergent validity was assessed using correlations
where values of 0.95 or greater indicate good fit; Root
between total CAT-Q and factor scores, and measures of
Mean Square Error of Approximation (RMSEA), where
autistic-like traits, social anxiety, well-being, generalised
values of 0.06 or lower indicate acceptable fit; and Stand-
anxiety, and depression.
ardised Root Mean Square Residual (SRMR), where

13
Journal of Autism and Developmental Disorders

Results These were reduced to twenty-five items based on the high-

est factor loadings, which resulted in a total of 8 items each
The characteristics of the total sample and all subsamples in the Masking and Assimilation factors, and 9 items in the
are described in Table 1. Compensation factor. Table 3 presents the mean scores and
internal consistencies of the factors and total scale across
Exploratory Analyses the autistic, non-autistic, and combined samples. Autistic
participants scored significantly higher than non-autistic
Parallel analysis suggested four factors, but examination of participants on the Total CAT-Q (t [401] = 12.98, p < .001;
the scree plot and eigenvalues suggested that three common partial η 2 = 0.30) and Compensation (t [401] = 11.90,
factors best fit the data across the autistic, non-autistic, and p < .001; partial η2 = 0.26), Masking (t [401] = 2.19, p = .03;
combined samples, in addition to being a simpler structure. partial η2 = 0.01), and Assimilation factors (t [401] = 16.35,
The three factors were labelled Compensation (strategies p < .001; partial η2 = 0.40). Factor loadings on all three fac-
used to actively compensate for difficulties in social situa- tors in the final, 25-item Camouflaging Autistic Traits Ques-
tions), Masking (strategies used to hide autistic character- tionnaire (CAT-Q) are detailed in Table 4.
istics or portray a non-autistic persona), and Assimilation
(strategies that reflect trying fit in with others in social situ- Confirmatory Analyses
ations). These three factors accounted for 38% of variance
in the autistic subsample, 41% of variance in the non-autistic Confirmatory factor analysis was performed on the con-
subsample, and 45% of variance in the combined exploratory firmatory sample (N = 419; Autistic N = 150, Non-Autis-
sample. Factor correlations were medium-to-high (Cohen tic N = 269); the results for the autistic, non-autistic, and
1988) between all factors in all samples (Table 2). combined group analyses for the total scale are reported in
Items that loaded onto one of the three factors at or above Table 5.
the critical value of 0.40 in both the autistic and non-autistic Overall the model fit was acceptable; CFI values were
subsamples, and in the combined sample, were identified. above 0.95, and RMSEA and SRMR values were well within

Table 1  Sample characteristics
Total sample Autistic subsample Non-autistic subsample Exploratory subsample Confirmatory subsample

N 832 354 478 402 430

(male/female/other gender/ (300/434/46/52) (108/179/17/50) (192/255/29/2) (139/246/17/0) (161/188/29/52)
not stated)
Mean age in years (SD) 36.01 41.93 30.24 37.02 35.15
(14.84) (13.55) (13.72) (15.02) (14.21)
Age range in years 16–82 18–75 16–82 16–82 16–72
Mean age at autism diagno- – 34.2 – 34.47a 33.82a
sis (range) (2–66) (2–66) (3–66)
Native language = English 617 244 373 346 271
Employed full- or part-time 308 135 173 182 126
Student 257 36 221 123 134
Retired or homemaker 62 43 19 39 23
Unemployed or unable to 86 64 22 42 44
work

NB: some participants chose not to answer some demographic questions

a
 Value for autistic participants only

Table 2  Factor correlations Compensation Masking Assimilation

in autistic, non-autistic and
combined (Com) samples Autistic Non-Autistic Com Autistic Non-autistic Com Autistic Non-autistic Com

Compensation – – – 0.5 0.47 0.39 0.44 0.58 0.66

Masking 0.5 0.47 0.39 – – – 0.21 0.39 0.33
Assimilation 0.44 0.58 0.66 0.21 0.39 0.33 – – –

13
 Journal of Autism and Developmental Disorders

Table 3  CAT-Q total and factor Scale No. of items Mean (SD) Internal consistency (Cron-
scores in the autistic (N = 200) bach’s α)
and non-autistic (N = 202)
subsamples and the combined Autistic Non-autistic Com Autistic Non-autistic Com
exploratory sample (Com;
N = 402) Total 25 4.79 (0.99) 3.48 (1.04) 4.13 (1.21) 0.91 0.93 0.94
Compensation 9 4.42 (1.31) 2.89 (1.27) 3.65 (1.50) 0.88 0.90 0.92
Masking 8 4.55 (1.35) 4.29 (1.10) 4.42 (1.24) 0.87 0.84 0.86
Assimilation 8 5.29 (1.15) 3.32 (1.27) 4.30 (1.56) 0.86 0.89 0.93

Raw Scores have been rescaled to reflect the 7-Point Likert Scale

the recommended range in all three samples. The model Compensation factor, while moderate stability was found for
tested is detailed in Fig. 1. the Masking and Assimilation factors (Table 7). No signifi-
cant difference between scores at Time 1 and Time 2 was
Invariance Analyses found (F[1, 29] = 0.23, p = .63).

Measurement invariance (as demonstrated by ∆CFI < 0.01) Validation

was found for item loadings (Model 2), intercepts (Model 3),
and residuals (Model 4) between the autistic and non-autistic Correlations were performed between the total and factor
male and female samples (as shown in Table 6). Model fit CAT-Q scores, and scores on autistic-like traits (total BAPQ
was close to or within acceptable limits for all models. It was score and subscale scores), social anxiety (total LSAS
concluded that the CAT-Q demonstrates strict invariance score), wellbeing (total WEMWBS score), generalised
between autistic and non-autistic males and females. anxiety (total GAD-7 score), and depression (total PHQ-9
score) in order to investigate convergent validity. Results in
Reliability and Validity the autistic and non-autistic samples are detailed in Table 8.
Generalised anxiety and depression scores were available for
Reliability and validity of the finalised, 25-item scale were autistic participants only.
assessed in a subset of the total sample that had also pro- The total CAT-Q score and all CAT-Q factors were sig-
vided complete responses to at least one of the other meas- nificantly positively correlated with autistic-like traits and
ures included (N = 706; Autistic N = 306, Non-Autistic social anxiety in autistic and non-autistic samples, with the
N = 400). exception of the Masking factor, which was not significantly
related to autistic-like traits in the autistic sample. The total
Reliability CAT-Q and all CAT-Q factors were significantly nega-
tively correlated with wellbeing in the non-autistic sample;
High internal consistency was found for the total scale however, in the autistic sample, only total CAT-Q and the
(Cronbach’s α = 0.94), and the Compensation (α = 0.91), Assimilation factor were significantly negatively correlated
Masking (α = 0.85), and Assimilation (α = 0.92) factors. with wellbeing. Depression and generalised anxiety were
Correlations between each factor and the total score were only measured in the autistic sample; both of these were
calculated using item-total correlation, and the corrected significantly positively correlated with total CAT-Q and all
factor-total correlations in the total sample were: Compen- its factors.
sation α = 0.705; Masking α = 0.483, Assimilation α = 0.627.
Test–retest reliability was calculated in a subsample of
30 autistic participants, who completed the CAT-Q again Discussion
online three months after initial testing. This subsample was
significantly older on average than the total autistic sample This study psychometrically tested the newly developed
(F[1, 331] = 12.61, p < .001; mean difference = 9.23 years Camouflaging Autistic Traits Questionnaire (CAT-Q) in
[SE = 2.6]). However, there was no significant difference autistic and non-autistic samples. Exploratory factor analysis
in the distribution of genders (Male, Female, Other Gen- identified a three-factor structure, consisting of Compensa-
der, and not stated) (Χ2 [4] = 1.66, p = .80), and no signifi- tion (strategies used to compensate for social and commu-
cant difference in mean Total BAPQ score (t[299] = 0.55, nication difficulties), Masking (strategies used to present a
p = .59) between this subsample and the total autistic sam- non-autistic or less autistic persona to others), and Assimi-
ple. Good stability was found, as measured by Pearson’s r lation (strategies used to fit in to uncomfortable social situ-
and intra-class correlations (ICC) for the total scale and the ations). The structure of the refined, 25-item CAT-Q (see

13
Journal of Autism and Developmental Disorders

Table 4  Factors loadings of the 25-Item CAT-Q in autistic, non-autistic and combined (Com) exploratory subsamples
Item Factors
Compensation Masking Assimilation
Autistic Non-autistic Com Autistic Non-autistic Com Autistic Non-autistic Com

When I am interacting with someone, I delib- 0.48 0.60 0.58 0.17 0.08 0.15 0.00 0.06 0.01
erately copy their body language or facial
expressions
I learn how people use their bodies and faces to 0.73 0.76 0.77 0.09 0.04 0.06 – 0.10 0.01 0.02
interact by watching television or films, or by
reading fiction
I have tried to improve my understanding of 0.73 0.73 0.73 0.04 0.03 0.04 – 0.13 0.02 0.04
social skills by watching other people
I will repeat phrases that I have heard others say 0.59 0.53 0.57 – 0.28 0.15 – 0.06 0.22 0.09 0.18
in the exact same way that I first heard them
I practice my facial expressions and body lan- 0.51 0.61 0.61 0.32 0.15 0.25 – 0.03 0.04 – 0.03
guage to make sure they look natural
I have spent time learning social skills from tel- 0.83 0.72 0.79 – 0.01 – 0.01 0.01 – 0.03 0.15 0.09
evision shows and films, and try to use these
in my interactions
In my own social interactions, I use behaviours 0.76 0.74 0.73 0.02 0.13 0.10 – 0.03 0.01 0.04
that I have learned from watching other peo-
ple interacting
I have researched the rules of social interactions 0.61 0.41 0.56 0.06 0.02 0.03 – 0.01 0.15 0.14
(for example, by studying psychology or read-
ing books on human behaviour) to improve
my own social skills
I have developed a script to follow in social 0.53 0.46 0.48 0.02 0.17 0.09 0.28 0.16 0.28
situations (for example, a list of questions or
topics of conversation)
I monitor my body language or facial expres- 0.03 0.32 0.17 0.85 0.60 0.75 0.02 0.04 – 0.02
sions so that I appear relaxed
I adjust my body language or facial expressions 0.08 0.31 0.22 0.79 0.56 0.69 – 0.06 0.06 – 0.04
so that I appear relaxed
I monitor my body language or facial expres- 0.11 0.22 0.16 0.71 0.52 0.66 0.12 0.09 0.03
sions so that I appear interested by the person
I am interacting with
I adjust my body language or facial expressions 0.06 0.23 0.15 0.74 0.57 0.69 0.07 0.08 0.02
so that I appear interested by the person I am
interacting with
I don’t feel the need to make eye contact with – 0.02 – 0.23 – 0.18 0.59 0.30 0.52 – 0.01 0.14 0.01
other people if I don’t want to (Reversed)
In social interactions, I do not pay attention to 0.00 0.07 0.03 0.82 0.47 0.69 – 0.11 0.12 – 0.01
what my face or body are doing (Reversed)
I always think about the impression I make on – 0.01 0.02 – 0.08 0.39 0.61 0.52 0.17 0.05 0.13
other people
I am always aware of the impression I make on – 0.10 – 0.06 – 0.16 0.44 0.63 0.54 0.01 – 0.16 – 0.08
other people
I rarely feel the need to put on an act in order to 0.00 – 0.13 – 0.08 0.24 0.19 0.21 0.56 0.69 0.71
get through a social situation (Reversed)
When talking to other people, I feel like the – 0.08 – 0.11 – 0.03 – 0.14 – 0.02 – 0.13 0.70 0.75 0.85
conversation flows naturally (Reversed)
When in social situations, I try to find ways to 0.01 0.28 0.14 – 0.21 – 0.21 – 0.18 0.66 0.66 0.75
avoid interacting with others
In social situations, I feel like I’m “performing” 0.04 0.04 0.06 0.12 0.27 0.11 0.70 0.57 0.75
rather than being myself
I have to force myself to interact with people 0.06 0.17 0.10 – 0.05 – 0.11 – 0.05 0.72 0.72 0.77
when I am in social situations

13
 Journal of Autism and Developmental Disorders

Table 4  (continued)
Item Factors
Compensation Masking Assimilation
Autistic Non-autistic Com Autistic Non-autistic Com Autistic Non-autistic Com

In social situations, I feel like I am pretending 0.00 0.21 0.09 0.19 0.18 0.13 0.65 0.58 0.74
to be “normal”
I need the support of other people in order to 0.08 0.31 0.16 – 0.11 – 0.04 – 0.07 0.60 0.52 0.66
socialise
I feel free to be myself when I am with other – 0.07 – 0.15 – 0.10 0.19 0.10 0.09 0.63 0.69 0.81
people (Reversed)

Loadings of 0.30 and greater are in bold

Online Appendix 1) was corroborated through confirmatory in situations where they do not know others well, whereas
factor analysis, and measurement invariance was established they feel free to be themselves while alone or with trusted
between all four groups, suggesting that the CAT-Q is appro- others (Hull et al. 2017).
priate for use in clinical and non-clinical populations, and The model tested here provided a good fit in both autistic
that scores can be compared between males and females. and non-autistic samples. Total CAT-Q score was positively
The CAT-Q demonstrated acceptable to good internal con- correlated with autistic-like traits in both samples, sug-
sistency and reliability over a period of 3 months. However, gesting that the higher level of autistic-like traits a person
as the test–retest reliability analyses were conducted only in has, the more they will camouflage those traits, regardless
the older autistic sample, we report these findings as prelimi- of autism diagnosis. As high-level, successful camouflag-
nary and suggest future research replicates these analyses in ing may result in missed clinical diagnoses (Tierney et al.
more diverse autistic and non-autistic samples. 2016), the CAT-Q could be used to identify camouflaging
The factors of Compensation and Masking reflect the behaviours in individuals considered at-risk for autism, but
two components of camouflaging proposed in a previous who do not currently meet diagnostic criteria. Measurement
conceptual model derived from qualitative research (Hull invariance analyses also demonstrated that the underlying
et al. 2017). The third factor (‘Assimilation’) represents structure of the CAT-Q is comparable in male and female
attempts to blend in to social situations in which the indi- autistic and non-autistic samples; in other words, the CAT-Q
vidual is uncomfortable, without letting others see this measures the same latent constructs in both genders and
discomfort. These motivations for camouflaging have been diagnostic groups. However, autistic participants scored
described in previous research, although not extensively significantly higher than non-autistic participants on the
(Hull et al. 2017; Tint and Weiss 2017). The strategies total CAT-Q and all three factors in the exploratory sample,
within the Assimilation factor included avoiding social sit- demonstrating that the CAT-Q measures behaviours that are
uations or managing them with the help of others, along- more common in individuals who have been diagnosed with
side items reflecting the feeling of not being one’s self autism spectrum conditions.
during interactions. The factor reflects comments made The Masking factor demonstrated the smallest difference
by autistic adults that they often choose to camouflage between autistic and non-autistic samples in this analysis,
suggesting that there may be more overlap between these
two groups than for the other factors. Masking may be less
Table 5  Fit of 25-Item Full CAT-Q Scale across autistic (N = 154),
non-autistic (N = 276) and Combined confirmatory samples (Com; specific to autism than the other components of camou-
N = 430) flaging, and may reflect more general self-presentation or
Sample Χ2 Df CFI RMSEA (90% CI) SRMR
impression-management strategies applied to autistic char-
acteristics. However, further research is needed to directly
Autistic 596.947* 272 0.970 0.056 (0.050–0.063) 0.075 compare masking strategies and other self-presentation
Non-Autistic 619.099* 272 0.983 0.046 (0.041–0.051) 0.058 strategies in autistic and non-autistic samples to determine
Com 969.527* 272 0.980 0.052 (0.048–0.055) 0.057 similarities and differences. In the autistic sample, mask-
Robust Statistics are reported
ing was not significantly correlated with autistic-like traits,
suggesting that it may be a response to the identification of
CFI Comparative Fit Index, RMSEA root mean square error of
approximation, 90% CI 90% confidence intervals, SRMR standardised being autistic rather than to the presence of specific autistic
root mean square residual characteristics; in contrast, a significant positive relationship
*p < .0001. Χ2 = Chi squared; Df = degrees of freedom between the two was observed for the non-autistic sample,

13
Journal of Autism and Developmental Disorders

Fig. 1  Social Camouflaging
model

Table 6  Multi-group measurement invariance model comparison

(autistic male N = 107; autistic female N = 181; non-autistic male
N = 192; non-autistic female N = 256)

Model Χ2 ∆Χ2 Df CFI ∆CFI Table 7  Test–retest reliability of Camouflaging Autistic Traits Ques-
tionnaire and factors in autistic subsample (N = 30)
1. Configural invariance 2434.22 – 1088 0.947 –
Pearson’s r ICC[C,1] 95% CI
2. Metric invariance 2353.93 80.29 1154 0.953 0.006
3. Scalar invariance 2628.31 272.38 1220 0.945 0.008 Total CAT-Q 0.77 0.77 0.73, 0.79
4. Residual invariance 2856.15 227.84 1295 0.939 0.006 Compensation factor 0.78 0.77 0.72, 0.82
Masking factor 0.70 0.70 0.63, 0.76
Robust statistics are reported
Assimilation factor 0.73 0.73 0.67, 0.78
Χ2 chi squared, ∆Χ2 chi square difference, Df degrees of freedom,
CFI Comparative Fit Index, ∆CFI CFI difference

13
 Journal of Autism and Developmental Disorders

Table 8  Correlations between CAT-Q Total and factor scores and autistic traits (BAPQ), social anxiety (LSAS), wellbeing (WEMWBS), depres-
sion (PHQ), and generalised anxiety (GAD) for the autistic (N = 306) and non-autistic (N = 400) subsamples
Total BAPQ BAPQ: Aloof BAPQ: prag- BAPQ: rigidity Total LSAS WEMWBS PHQ GAD
matic language

Autistic
 CAT-Q total 0.34*** 0.24*** 0.33*** 0.28*** 0.44*** − 0.16* 0.28*** 0.35***
 Compensation 0.21*** 0.08 0.27*** 0.18** 0.30*** − 0.02 0.18** 0.25***
 Masking − 0.03 − 0.07 − 0.03 0.01 0.19** − 0.02 0.16** 0.20***
 Assimilation 0.72*** 0.63*** 0.62*** 0.54*** 0.60*** − 0.37*** 0.35*** 0.41***
Non-autistic
 CAT-Q total 0.67*** 0.58*** 0.56*** 0.54*** 0.60*** − 0.43*** – –
 Compensation 0.54*** 0.42*** 0.52*** 0.44*** 0.46*** − 0.31*** – –
 Masking 0.32*** 0.24*** 0.24*** 0.32*** 0.35*** − 0.24*** – –
 Assimilation 0.78*** 0.77*** 0.62*** 0.59*** 0.69*** − 0.53*** – –

*p < .05, **p < .01, ***p < .001

suggesting that the two groups may have been using masking autism diagnostic measures, to enhance the sensitivity and
strategies in response to different motivations. specificity of clinical diagnosis, formulation, and support
Previous research suggested that camouflaging in autistic planning; however, the clinical utility requires further clini-
adults may be associated with poor mental health outcomes, cal research to establish.
especially anxiety, depression, and generally poor quality of In addition, the CAT-Q does not require an official
life (Cage et al. 2017; Hull et al. 2017; Lai et al. 2017). The diagnosis of an autism spectrum condition for camouflag-
positive correlations between the CAT-Q and measures of ing behaviours to be assessed, as the underlying structure
social anxiety, anxiety, and depression, and the negative cor- shows invariance between autistic and non-autistic popula-
relation between the CAT-Q and wellbeing, support this idea tions. This addresses some issues in current autism research,
and offer convergent validation of the measure. Greater total especially that criteria for autistic participants may be based
camouflaging appears to be associated with poorer mental on an overly restricted and potentially inaccurate opera-
health outcomes overall, although interestingly the Compen- tional definition of autism. Even if autism diagnostic crite-
sation and Masking factors were not significantly associated ria change in the future, use of the CAT-Q should not vary
with wellbeing in the autistic sample. This may reflect indi- between clinical and non-clinical groups. The CAT-Q has
vidual differences in the impact or success of camouflaging; demonstrated measurement invariance between male and
previous research found that associations between camou- female participants, enabling comparison across genders in
flaging and negative outcomes were stronger for autistic men future research.
than women (Lai et al. 2017). Further assessment of gender This study is not without its limitations. First, although
differences and other individual differences in camouflaging the BAPQ has demonstrated validity and reliability in clini-
behaviours and their association with wellbeing and mental cal and non-clinical samples (Ingersoll et al. 2011; Nishiy-
health in this sample is currently underway. ama et al. 2014), it was developed for use with relatives of
those with an autism diagnosis. Therefore we are cautious
Strengths and Limitations about using BAPQ scores as a measure of autistic traits in
clinical and general population samples (Piven and Sasson
A significant strength of this approach is that the items were 2014). In future, to accurately examine how camouflaging is
developed based on information from autistic people them- related to autistic traits, the CAT-Q should be compared to
selves, describing their own experiences of camouflaging. a measure of autistic traits which has been explicitly devel-
This ensures that behaviours which may not have been previ- oped for use in autistic populations, for example, the severity
ously identified as part of social camouflaging by non-autis- score of the ADOS-2.
tic clinicians and researchers can be measured. The CAT-Q Second, no behavioural measure of social ability was
can be used in combination with observed behavioural and included in the study. Individuals with greater social skills
cognitive measures of camouflaging to assess all aspects of are less likely to need to camouflage in the first place, and
this complex phenomenon. It may also have clinical impli- may do so more effectively than those with poorer social
cations to identify levels of camouflaging along with other skills. Further research is needed to identify the extent to
clinical information, including those derived from current which social skills predict camouflaging behaviours, which

13
Journal of Autism and Developmental Disorders

will have implications regarding prevailing social skills behaviours and compare between groups; in clinical set-
training in autistic individuals. There was also no objec- tings as a potential screening tool for individuals who may
tive validation of self-reported autism diagnosis. However, be missed under current autism diagnostic criteria because
only participants who reported receiving a diagnosis from a they camouflage; and by autistic and non-autistic people to
healthcare professional were included in the autistic sample. aid identification of beneficial or harmful behaviours they
Third, responses on the PHQ-9 and GAD-7 were not avail- use in social situations. Further validation of the CAT-Q in
able for non-autistic participants as these data were collected more diverse samples is encouraged in the future, alongside
as part of a separate project; the relationship between cam- comparison with existing measures of camouflaging and
ouflaging and depression and anxiety should therefore also broader social skills.
be examined in non-autistic adults.
Fourthly, the self-report CAT-Q only measures indi- Acknowledgements  The authors would like to thank all those who
contributed to the design of the questionnaire, who took part in the
viduals’ own reflections/perceptions of their camouflaging study, and who provided comments on the findings. The authors also
behaviours, and is thus limited in its use to those who are thank Maya Bowri, Emogen Campbell, Andrew Dunlop, and Lily
able to reflect on their own behaviours and provide insight Levy for assistance with data collection. During the period of the
to their motivations. The CAT-Q may therefore not be useful study Meng-Chuan Lai was supported by the University of Toronto
Department of Psychiatry Excellence Fund and the O’Brien Scholars
for autistic individuals with language difficulties or intellec- Program within the Child and Youth Mental Health Collaborative at
tual disability. By combining this measure with behavioural the Centre for Addiction and Mental Health and The Hospital for Sick
or informant-report measures of camouflaging, estimates of Children, Toronto, Canada; Simon Baron-Cohen was supported by the
camouflaging behaviours in those who have less insight or Autism Research Trust, Autistica, the MRC and Wellcome Trust, and
by the National Institute for Health Research (NIHR) Collaboration for
ability to communicate it can also be obtained. Leadership in Applied Health Research and Care East of England at
Fifthly, the CAT-Q was created mainly based on reflec- Cambridgeshire and Peterborough NHS Foundation Trust. The views
tions from autistic adults, and was psychometrically exam- expressed are those of the author(s) and not necessarily those of the
ined and validated in the present adult sample, in which a NHS, the NIHR or the Department of Health.
substantial proportion of the autistic participants received
Author Contributions  LH conceived of the study, participated in the
their diagnoses in adulthood instead of childhood. Hence, design and coordination of the study, performed analyses, and drafted
although the validity and potential clinical utility are likely the manuscript. WM, MCL, and SBC conceived of the study, par-
ensured in autistic adults, in particular those who are diag- ticipated in its design, coordination, and analyses, and revised the
nosed in adulthood (Lai and Baron-Cohen 2015), it is still manuscript. CA and PS conceived of the study, and participated in its
coordination. KVP participated in the design and analyses, and revised
unclear whether the psychometric properties and poten- the manuscript. All authors read and approved the final manuscript.
tial utilities hold for adolescents and older children, with
or without autism, or for those with intellectual disability. Compliance with Ethical Standards 
Further testing of the CAT-Q in samples of varying ages and
abilities, including adults who were diagnosed in childhood, Conflict of interest  The authors declare that they have no conflict of
should be conducted to measure its factor structure, valid- interest.
ity and reliability across these groups. As the confirmatory
Ethical Approval  All procedures performed in studies involving human
sample contained more males than females, these analy- participants were in accordance with the ethical standards of the insti-
ses should also be replicated in a gender-matched sample. tutional and/or national research committee and with the 1964 Helsinki
Finally, although the validation of the CAT-Q supports declaration and its later amendments or comparable ethical standards.
previous research suggesting camouflaging is associated
Informed Consent  Informed consent was obtained from all individual
with poorer wellbeing and mental health outcomes, only participants included in the study.
correlational relationships were identified. Longitudinal or
intervention researches are necessary to confirm the causal
Open Access  This article is distributed under the terms of the Crea-
nature of these relationships, and to establish the mecha- tive Commons Attribution 4.0 International License (http://creat​iveco​
nisms and individual characteristics that may predict out- mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribu-
comes of camouflaging. tion, and reproduction in any medium, provided you give appropriate
credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made.

Conclusions

The CAT-Q is a valid and reliable self-report measure of

adults’ social camouflaging behaviours, suitable for use in
autistic and non-autistic male and female populations. It
can be used in research settings to quantify camouflaging

13
 Journal of Autism and Developmental Disorders

References camouflaging in adults with autism spectrum conditions. Journal

of Autism and Developmental Disorders, 47(8), 2519–2534. https​
://doi.org/10.1007/s1080​3-017-3166-5.
American Psychiatric Association (2013). DSM 5. Washington: Ameri-
Hurley, R. S. E., Losh, M., Parlier, M., Reznick, J. S., & Piven, J.
can Psychiatric Association.
(2007). The broad autism phenotype questionnaire. Journal of
Baker, S. L., Heinrichs, N., Kim, H., & Hofmann, S. G. (2002). The
Autism and Developmental Disorders, 37(9), 1679–1690. https​://
Liebowitz Social Anxiety Scale as a self-report instrument: A pre-
doi.org/10.1007/s1080​3-006-0299-3.
liminary psychometric analysis. Behaviour Research and Therapy,
Ingersoll, B., Hopwood, C. J., Wainer, A., & Brent Donnellan, M.
40, 701–715. https​://doi.org/10.1016/S0005​-7967(01)00060​-2.
(2011). A comparison of three self-report measures of the broader
Bargiela, S., Steward, R., & Mandy, W. (2016). The experiences of
autism phenotype in a non-clinical sample. Journal of Autism
late-diagnosed women with autism spectrum conditions: An
and Developmental Disorders, 41(12), 1646–1657. https​://doi.
investigation of the female autism phenotype. Journal of Autism
org/10.1007/s1080​3-011-1192-2.
and Developmental Disorders, 46(10), 3281–3294. https​://doi.
Kenny, L., Hattersley, C., Molins, B., Buckley, C., Povey, C., &
org/10.1007/s1080​3-016-2872-8.
Pellicano, E. (2015). Which terms should be used to describe
Begeer, S., Mandell, D., Wijnker-Holmes, B., Venderbosch, S., Rem,
autism? Perspectives from the UK autism community. Autism.
D., Stekelenburg, F., & Koot, H. M. (2013). Sex differences in the
https​://doi.org/10.1177/13623​61315​58820​0.
timing of identification among children and adults with autism
Koo, T. K., & Li, M. Y. (2016). A guideline of selecting and report-
spectrum disorders. Journal of Autism and Developmental Dis-
ing intraclass correlation coefficients for reliability research.
orders, 43, 1151–1156.
Journal of Chiropractic Medicine, 15(2), 155–163. https​://doi.
Cage, E., Di Monaco, J., & Newell, V. (2017). Experiences of autism
org/10.1016/j.jcm.2016.02.012.
acceptance and mental health in autistic adults. Journal of Autism
Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9.
and Developmental Disorders, 48, 473–484.
Journal of General Internal Medicine, 16, 606–613.
Cage, E., Pellicano, E., Shah, P., & Bird, G. (2013). Reputation man-
Lai, M.-C., & Baron-Cohen, S. B. (2015). Identifying the lost gen-
agement: Evidence for ability but reduced propensity in autism.
eration of adults with autism spectrum conditions. The Lancet
Autism Research, 6(5), 433–442. https:​ //doi.org/10.1002/aur.1313.
Psychiatry, 2(11), 1013–1027.
Cattell, R. B. (1966). The scree test for the numbers of factors. Multi-
Lai, M.-C., Lombardo, M. V., Pasco, G., Ruigrok, A. N. V., Wheel-
variate Behavioral Research, 1, 245–276. https:​ //doi.org/10.1207/
wright, S. J., Sadek, S. A., … Baron-Cohen, S. (2011). A behav-
s1532​7906m​br010​2.
ioral comparison of male and female adults with high function-
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit
ing autism spectrum conditions. PLoS ONE, 6(6), e20835. https​
indexes for testing measurement invariance. Structural Equation
://doi.org/10.1371/journ​al.pone.00208​35.
Modeling: A Multidisciplinary Journal, 9(2), 233–255. https​://
Lai, M.-C., Lombardo, M. V., Ruigrok, A. N. V., Chakrabarti, B.,
doi.org/10.1207/S1532​8007S​EM090​2.
Auyeung, B., Szatmari, P., … Consortium, M. A. (2017). Quan-
Cohen, J. (1988). Set correlation and contingency tables. Applied
tifying and exploring camouflaging in men and women with
Psychological Measurement, 12(4), 425–434. https ​ : //doi.
autism. Autism, 21(6), 690–702. https​://doi.org/10.1177/13623​
org/10.1177/01466​21688​01200​410.
61316​67101​2.
Constantino, J. N. (2011). The quantitative nature of autistic social
Leary, M. R., & Kowalski, R. M. (1990). Impression management:
impairment. Pediatric Research, 69(5 Pt 2), 55R–62R. https​://
A literature review and two-component model. Psychological
doi.org/10.1203/PDR.0b013​e3182​12ec6​e.
Bulletin, 107(1), 34–47.
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in pyscho-
Lehnhardt, F.-G., Falter, C. M., Gawronski, A., Pfeiffer, K., Tep-
logical tests. Psychological Bulletin, 52, 281–302.
est, R., Franklin, J., & Vogeley, K. (2015). Sex-related cog-
Dean, M., Harwood, R., & Kasari, C. (2017). The art of camouflage:
nitive profile in autism spectrum disorders diagnosed late in
Gender differences in the social behaviors of girls and boys with
life: Implications for the female autistic phenotype. Journal of
autism spectrum disorder. Autism, 21(6), 678–689. https​://doi.
Autism and Developmental Disorders, 46(1), 139–154. https​://
org/10.1177/13623​61316​67184​5.
doi.org/10.1007/s1080​3-015-2558-7.
DiStefano, C., & Morgan, G. B. (2014). A comparison of diagonal
Liebowitz, M. R. (1987). Social phobia. In M. R. Liebowitz (ed)
weighted least squares robust estimation techniques for ordinal
Anxiety (Vol. 22, pp. 141–173). Basel: Karger Publishers.
data. Structural Equation Modeling, 21(3), 425–438. https​://doi.
Livingston, L. A., & Happé, F. (2017). Conceptualising compensa-
org/10.1080/10705​511.2014.91537​3.
tion in neurodevelopmental disorders: Reflections from autism
Giarelli, E., Wiggins, L. D., Rice, C. E., Levy, S. E., Kirby, R. S.,
spectrum disorder. Neuroscience and Biobehavioral Reviews.
Pinto-Martin, J., & Mandell, D. (2010). Sex differences in the
https​://doi.org/10.1016/j.neubi​orev.2017.06.005.
evaluation and diagnosis of autism spectrum disorders among
Loomes, R., Hull, L., & Mandy, W. P. L. (2017). What is the male-
children. Disability and Health Journal, 3(2), 107–116. https​://
to-female ratio in autism spectrum disorder? A systematic
doi.org/10.1016/j.dhjo.2009.07.001.
review and meta-analysis. Journal of the American Academy
Hayton, J. C., Allen, D. G., & Scarpello, V. (2004). Factor retention
of Child & Adolescent Psychiatry, 56(6), 466–474. https​://doi.
decisions in exploratory factor analysis: A tutorial on parallel
org/10.1016/j.jaac.2017.03.013.
analysis. Organizational Research Methods, 7(2), 191–205. https​
Loth, E., Spooren, W., Ham, L. M., Isaac, M. B., Auriche-Benichou,
://doi.org/10.1177/10944​28104​26367​5.
C., Banaschewski, T., … Murphy, D. G. M. (2015). Identifica-
Head, A. M., McGillivray, J. A., & Stokes, M. A. (2014). Gender dif-
tion and validation of biomarkers for autism spectrum disor-
ferences in emotionality and sociability in children with autism
ders. Nature Reviews Drug Discovery, 15(1), 70–73. https​://
spectrum disorders. Molecular Autism, 5(1), 19.
doi.org/10.1038/nrd.2015.7.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covari-
McGraw, K. O., & Wong, S. P. (1996). Forming inferences about
ance structure analysis: Conventional criteria versus new alterna-
some intraclass correlations coefficients. Psychological Meth-
tives. Structural Equation Modeling: A Multidisciplinary Journal,
ods, 1(1), 30–46. https​://doi.org/10.1037/1082-989X.1.4.390.
6(1), 1–55. https​://doi.org/10.1080/10705​51990​95401​18.
Nishiyama, T., Suzuki, M., Adachi, K., Sumi, S., Okada, K., Kishino,
Hull, L., Petrides, K. V., Allison, C., Smith, P., Baron-Cohen, S., Lai,
H., … Kanne, S. M. (2014). Comprehensive comparison of
M., & Mandy, W. (2017). "Putting on my best normal”: Social
self-administered questionnaires for measuring quantitative

13
Journal of Autism and Developmental Disorders

autistic-like traits in adults. Journal of Autism and Develop- Prevalence and diagnostic classification. Autism Research, 6(2),
mental Disorders, 44(5), 993–1007. https​: //doi.org/10.1007/ 134–143. https​://doi.org/10.1002/aur.1272.The.
s1080​3-013-2020-7. Skuse, D. H., Mandy, W. P. L., & Scourfield, J. (2005). Measuring
Parish-Morris, J., Liberman, M. Y., Cieri, C., Herrington, J. D., autistic-like traits: Heritability, reliability and validity of the
Yerys, B. E., Bateman, L., … Schultz, R. T. (2017). Linguistic Social and communication disorders checklist. The British Jour-
camouflage in girls with autism spectrum disorder. Molecular nal of Psychiatry, 187(6), 568–572. https​://doi.org/10.1192/
Autism, 8(1), 48. https​://doi.org/10.1186/s1322​9-017-0164-6. bjp.187.6.568.
Piven, J., & Sasson, N. J. (2014). On the misapplication of the broad Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A
autism phenotype questionnaire in a study of autism. Journal of brief measure for assessing generalized anxiety disorder. Archives
Autism and Developmental Disorders, 44(8), 2077–2078. https​ of Internal Medicine, 166(10), 1092. https:​ //doi.org/10.1001/archi​
://doi.org/10.1007/s1080​3-014-2076-z. nte.166.10.1092.
R Core Team. (2013). R: A language and environment for statistical Tennant, R., Hiller, L., Fishwick, R., Platt, S., Joseph, S., Weich, S.,
processing. Vienna: R Foundation for Statistical Computing. … Stewart-Brown, S. (2007). The Warwick–Edinburgh Mental
Ratto, A. B., Kenworthy, L., Yerys, B. E., Bascom, J., Trubanova, A., Well-being Scale (WEMWBS): Development and UK valida-
White, S. W., … Anthony, G. (2017). What About the Girls ? tion. Health and Quality of Life Outcomes, 5(1), 63. https​://doi.
Sex-Based Differences in Autistic-like traits and Adaptive Skills. org/10.1186/1477-7525-5-63.
Journal of Autism and Developmental Disorders. https​://doi. Thurmond, V. A. (2001). The point of triangulation. Journal of
org/10.1007/s1080​3-017-3413-9. Nursing Scholarship, 33(3), 253–258. https​://doi.org/10.111
Rutherford, M., McKenzie, K., Johnson, T., Catchpole, C., O’Hare, 1/j.1547-5069.2001.00253​.x.
A., McClure, I., … Murray, A. (2016). Gender ratio in a clinical Tierney, S., Burns, J., & Kilbey, E. (2016). Looking behind the
population sample, age of diagnosis and duration of assessment in mask: Social coping strategies of girls on the autistic spectrum.
children and adults with autism spectrum disorder. Autism. https​ Research in Autism Spectrum Disorders, 23, 73–83. https​://doi.
://doi.org/10.1177/13623​61315​61787​9. org/10.1016/j.rasd.2015.11.013.
Ruzich, E., Allison, C., Smith, P., Watson, P., Auyeung, B., Ring, H., Tint, A., & Weiss, J. A. (2017). A qualitative study of the service expe-
& Baron-Cohen, S. (2015). Measuring autistic-like traits in the riences of women with autism spectrum disorder. Autism. https​://
general population: A systematic review of the autism-spectrum doi.org/10.1177/13623​61317​70256​1.
quotient (AQ) in a nonclinical population sample of 6,900 typical Trousselard, M., Steiler, D., Dutheil, F., Claverie, D., Canini, F.,
adult males and females. Molecular Autism, 6(1), 2. https​://doi. Fenouillet, F., … Franck, N. (2016). Validation of the Warwick–
org/10.1186/2040-2392-6-2. Edinburgh Mental Well-Being Scale (WEMWBS) in French
Rynkiewicz, A., Schuller, B., Marchi, E., Piana, S., Camurri, A., psychiatric and general populations. Psychiatry Research, 245,
Lassalle, A., & Baron-Cohen, S. (2016). An investigation of 282–290. https​://doi.org/10.1016/j.psych​res.2016.08.050.
the ‘female camouflage effect’ in autism using a computerized Wang, W. E. C., & Cunningham, E. G. (2005). Comparison of alterna-
ADOS-2 and a test of sex/gender differences. Molecular Autism, tive estimation methods in confirmatory factor analyses of the
7(10), 1. https​://doi.org/10.1186/s1322​9-016-0073-0. General Health Questionnaire. Psychological Reports, 97, 3–10.
Sass, D. A. (2011). Testing measurement invariance and comparing Zeldovich, L. (2017). New global diagnostic criteria manual mirrors
latent factor means within a confirmatory factor analysis frame- U.S. autism criteria. Spectrum: Autism Research News. Retrieved
work. Journal of Psychoeducational Assessment, 29(4), 347–363. from https​://spect​rumne​ws.org.
https​://doi.org/10.1177/07342​82911​40666​1.
Sasson, N. J., Lam, K. S. L., Childress, D., Parlier, M., Daniels, J. L.,
& Piven, J. (2013). The broad autism phenotype Questionnaire:

13