Measuring masking

Methods fall broadly into two approaches: questionnaires that ask people how much they camouflage, and discrepancy methods, which compare how autistic a person seems to an observer with a measure of their underlying traits or abilities.¹ The authors of a systematic review of 29 studies propose that the two capture different things: the effort people feel they make, and how they come across.¹ A meta-analysis of 15 studies found that autistic people scored higher than non-autistic people on the CAT-Q questionnaire overall and on two of its three parts, though only five or six studies made each comparison. Autistic women and girls scored higher than autistic men and boys on both the questionnaire and discrepancy measures. Results varied a lot between studies, and an exploratory analysis of few studies suggested that age affected some comparisons, mainly among non-autistic people.²

A second systematic review found 16 distinct tools, all at early stages of testing. Twelve were questionnaires or observation schemes, three were discrepancy methods and one mixed the two.³ The same paper compared three methods in 59 autistic 13 to 18-year-olds. The self-report questionnaire and a discrepancy score were only weakly related, and not significantly (r = 0.20, with a range of plausible values from slightly negative to moderate). A parent-report version of the questionnaire tracked both more closely (0.47 and 0.46), and none of the three was related to social skills.³

These tools carry much of the research on what masking costs. The review of 29 studies found preliminary support for a link between more self-reported camouflaging and worse mental health, in cross-sectional studies whose samples did not represent autistic people as a whole.¹ How well the questionnaires work bears directly on those findings.

The most used questionnaire, the CAT-Q, has 25 items in three subscales: compensation (working around social difficulties, for example with scripts), masking (hiding autistic traits) and assimilation (forcing oneself to fit in). It was built from autistic adults' accounts and tested online in 354 autistic and 478 non-autistic adults.⁴ Its items agreed well with each other (α = 0.94). Scores were fairly stable when 30 autistic participants, older on average than the rest, took it again about three months later (r = 0.77); the authors call this result preliminary.⁴ A 9-item short form, built from the same 832 adults, kept the three-part structure in a separate group of 80.⁵ A critical review of 389 studies rated the CAT-Q's reliability excellent and the evidence for its validity mixed.⁶

The questionnaire has been translated and retested by other teams, with mixed results. Its three-part structure was confirmed in 1,227 French adults, 646 German adolescents and 948 Iranians aged 15 to 50 (with acceptable fit), and a nearly identical structure emerged from an exploratory analysis of a representative sample of 972 US adults.^8,9,10,7 In a Swedish sample aged 10 to 83 the scores were reliable, and autistic people and autistic women scored higher as expected, but the three-part structure did not fit well and there were validity problems in children under 15.¹¹ In 614 Japanese adults the original structure did not fit, and the researchers derived a shorter version; in its autistic sample two of the subscales were unrelated to mental health or autistic traits, and the authors advise caution with subscale scores.¹² In 205 Taiwanese adolescents the items formed two groups rather than three.¹³

Comparing scores between groups, such as autistic men and women, assumes the questions work the same way in each group. Tests of this assumption are called measurement invariance. The CAT-Q's developers concluded that it held across autistic and non-autistic men and women.⁴ Zachary Williams read their analysis as mixed and possibly too small to detect differences, and noted that it gave no sizes for any differences in how items behaved. Williams argues that the question bears on whether camouflaging means the same thing in non-autistic people.¹⁴

Later studies split. In German adolescents, scores could be compared across sex and age groups, and girls scored higher than boys only in the autistic group.⁹ In Iranian adults the CAT-Q worked the same way in people with many and with few autistic traits, a comparison of traits rather than of diagnosis.¹⁰ In French adults, by contrast, the tests suggested that the items mean something different to autistic and non-autistic people; the authors could not say whether this reflects culture or a real difference in what camouflaging means for non-autistic people.⁸

Non-autistic people also score on the CAT-Q. In a representative US sample the questionnaire had nearly the same structure as in samples with many autistic people, and higher scores went with comparing oneself to others, self-consciousness, internalized stigma and social anxiety.⁷ In 247 non-autistic young adults, those with lower social competence scored higher.¹⁵ In 254 adults, mostly women, diagnosed with autism, social anxiety disorder or both, total scores did not differ between the three groups; the author concludes that camouflaging is shared across diagnoses and driven by anxiousness and negative self-perception rather than by being autistic.¹⁶ In a Swedish sample, scores were lower in older non-autistic adults but not in older autistic adults; each person was measured once, so this shows differences between age groups, not change within a person.¹¹

Discrepancy methods treat camouflaging as a gap between how a person comes across and what their traits or test results would predict. The first study, of 60 autistic adults (30 women, 30 men) without intellectual disability, compared how autistic each person seemed to a clinician during the ADOS assessment with their self-reported traits and their score on a test of reading emotions from photographs of eyes. Women had larger gaps than men on average (d = 0.98), with wide variation in both groups.¹⁷ A study of 136 autistic adolescents compared ADOS scores with theory-of-mind tasks. Those who came across well despite low task scores had higher IQ and executive function, and reported more anxiety, than those who did not.¹⁸

Smaller studies have used other versions. A partial replication in 17 autistic men and 11 autistic women found camouflaging more common in the women.¹⁹ In 84 children and adolescents, 40 of them autistic or with many autistic traits (not all diagnosed), girls in that group showed more back-and-forth social behavior than boys with similar traits.²⁰ In 388 autistic children and adolescents, teachers rated girls as less autistic than boys while parents did not, and older age and being female went with larger parent-teacher gaps. The authors suggest the gap may reflect girls changing their behavior at school or teachers' expectations of girls.²¹

A critical review questions how well these gap scores have been validated.⁶ Williams notes, following an earlier critic, that some ADOS items rate insight and the expression of one's own feelings rather than outward behavior. Williams adds that non-autistic people score near the bottom of the ADOS, so their gaps can barely be measured, and that lab tests of social understanding relate only very modestly to real conversations.¹⁴ In a Swedish twin sample (315 twins, 69 of them autistic), a gap score did not correlate with the CAT-Q overall (r = 0.04), though it did moderately in the autistic twins (r = 0.41) and in adults (r = 0.27). Younger twins' traits were reported by parents rather than by themselves, so the gap score was not built the same way at every age.²²

Some researchers observe camouflaging directly. Watching 96 children at recess (48 autistic, 24 of them girls), one team saw autistic girls stay close to other children and weave in and out of their activities, which appeared to hide their social difficulties; autistic boys more often played alone. The autistic children had an IQ of 70 or above and attended mainstream classes.²⁴ In speech recorded during the ADOS from 65 autistic children (16 of them girls), autistic girls said "uh" less often than autistic boys, a pattern the authors call linguistic camouflage. The same difference between girls and boys appeared in a small group of 17 non-autistic children, so it may be an ordinary sex difference in filler words that also makes autistic girls sound less distinctive.²⁵ In a study of 93 autistic and non-autistic children, new conversation partners rated autistic girls more positively than autistic boys, though clinicians rated their autism as similar.²⁶

Other methods ask people about specific moments. In one study, 17 autistic adults were filmed talking with a stranger, then watched the recording and described where they had camouflaged.²⁷ A checklist of compensation strategies, given to 117 adults, found that people with more autistic traits and more education reported more strategies; this, rather than diagnosis itself, accounted for the higher scores of autistic participants. Scores were unrelated to sex or to age at diagnosis. Most of the sample were intellectually able women.²⁸ A 2026 study had 48 adults record video responses to two imagined situations, a job interview and a call with someone they trusted, always in the same order. Effort rose for everyone in the interview, but feeling inauthentic, monitoring oneself and anxiety rose more in the 23 autistic adults.²⁹ Williams has proposed comparing a person's ordinary conversation with one in which they are told to show their true self.¹⁴

Most measures were built for adults. Parent versions of the CAT-Q exist. In 59 autistic 13 to 18-year-olds, the parent version correlated with teenagers' own reports at 0.47.³ In 205 Taiwanese adolescents the self- and caregiver-report versions were highly correlated, though the items formed two groups rather than three,¹³ while in German adolescents agreement between teenagers and their parents was moderate in the autistic group and high in the non-autistic group.⁹ In the Swedish sample, the CAT-Q showed validity problems below age 15.¹¹

The Children's Masking Questionnaire, published in 2026, was developed together with families, young people and professionals. It has one part for an adult to answer and one for the child. As of October 2026, it has been evaluated only in two qualitative studies, by its developers, of whether people found it acceptable and appropriate.³⁰

Williams argues that research on what camouflaging goes with has outpaced work showing it is measured well. On this view, unless a camouflaging questionnaire can be told apart from social anxiety, its links with mental health in one-time surveys cannot be read, because the direction could run either way. Williams notes that different parts of camouflaging, including the CAT-Q subscales, have correlated only modestly in published studies, so a total score may not reflect one underlying trait. Self-report is open to biases such as agreeing with statements and wanting to look good, though Williams still considers it a promising way to measure camouflaging.¹⁴ Later studies have tested some of these points: a German study found the CAT-Q could be compared across sex and age in adolescents,⁹ and the factor study described below separated most items from social anxiety.³¹ Whether scores mean the same in autistic and non-autistic people remains unsettled.^8,10

Two studies of the CAT-Q point different ways. A critical review of 389 studies judged support for its validity mixed, arguing that scores are confounded by social anxiety and not specific to autism.⁶ A factor study of 308 autistic adults found most CAT-Q items separate from social anxiety, fear of negative evaluation and autistic social traits; the exception was the assimilation items, which overlapped mostly with autistic social traits.³¹

Who has been studied also limits what the measures show. A systematic review of 46 studies of passing as non-autistic, covering work up to May 2020, found that nearly all excluded people with an intellectual disability, and fewer than a quarter reported participants' race or ethnicity.³² The critical review found autistic women diagnosed in adulthood overrepresented, and people with intellectual or language disabilities or higher support needs underrepresented.⁶

Several autistic writers describe masking as hard to see, from inside and from outside. Megan Griffith writes that masking is "often not even a conscious decision, especially for those of us who started masking very young." She describes years of not recognizing it: "I definitely didn't know I was masking for a long time, even when others tried to tell me." She knows people who were aware of it from a young age, and others "who still aren't sure what's masking and what's not." One sign she names is an inner voice telling you to make eye contact or to move a certain way.^c1

Sol Smith writes that most people know little of who they are without their mask, because there was no particular day when they put it on. Smith suggests looking instead for "clues as to where you and your mask diverge," such as activities once enjoyed that later felt off limits.^c2 Smith describes masking as being "nearly obsessed with how you come across to others in every interaction," caught in a cycle of anticipating, rehearsing, watching and reviewing each exchange.^c4

Writers also describe observers misreading the surface. Before Smith's diagnosis, a doctor doubted it: "I don't know, you make pretty good eye contact." Smith writes of having "a long list of eye contact work-arounds."^c3 Devon Price writes that outgoing autistic people "are seldom correctly identified and diagnosed as kids," because they do not fit the picture of autism as cold and withdrawn.^c5 One autistic contributor to a clinicians' guide, identified as L, describes a psychologist who did not take them seriously because their emotions did not show, and now letting themselves "sink into" feeling unwell before appointments so that they cry and are believed.^c11

Amy Pearson, an autistic psychologist, and Kieran Rose, an autistic trainer and researcher, question both main ways of measuring masking. A gap between a clinician's rating and a person's own report, they argue, "does not necessarily indicate masking": assessments were built around a stereotyped picture of autism, the clinic is an artificial setting, and a person may leave out a strong interest simply because it seems beside the point. Questionnaires avoid that problem but depend on what people consciously know of their own masking, and some autistic people describe masking as not always under conscious control.^c6 Pearson and Rose argue for an idea of masking that covers all autistic people, "not just those who are explicitly aware, or can report their experiences fluently".^c7

Practitioners raise the same problem in the assessment room. Donna Henderson, Sarah Wayland and Jamell White, in a guide for clinicians, warn that many standard tests are less sensitive to autistic people who camouflage: "If I had to choose between interviewing and tests, I'd choose interviews every time!" They suggest using a person's answers on the CAT-Q to start a conversation about their social life.^c8,c9 Luke Beardon, a senior lecturer in autism, argues that a diagnostic assessment, with unfamiliar people, an unfamiliar task and little predictability, is just the kind of situation in which an anxious autistic adult may "almost automatically" fall into a "default masking persona".^c10