Fourteen late-diagnosed autistic women described what got missed

This 2016 paper is one of the studies most often cited on our pages about autism in women, late diagnosis, masking, and missed assessment. It helped move the conversation away from the question “Why don’t these women look autistic?” and toward a better one: “What were our referral and assessment systems trained to notice?”

It also shows why an influential qualitative study needs careful reading. The participants’ accounts are detailed and consequential. They generate patterns worth testing. They do not produce a checklist that can identify autism in someone who was not part of the study.

The researchers wanted to understand the experiences of women whose autism was not recognised in childhood. They asked how gender may have affected recognition, what strategies participants used to navigate social life, and how late diagnosis shaped identity and support.

The phrase “female autism phenotype” was the paper’s working frame. The study was designed to develop ideas about that frame, not to prove that autistic women form one uniform group or that the reported experiences occur only in women.

The team interviewed 14 women in the United Kingdom. Participants were 22 to 30 years old and had received an autism diagnosis at age 15 or later. All reported a diagnosis made by a qualified NHS professional and scored above the AQ-10 cut-off; the researchers did not retrieve clinical records. Everyone had an estimated IQ above 70.

Nine interviews happened in person, four by video, and one by phone. Offering remote formats reduced the chance that anxiety, sensory needs, or discomfort with face-to-face interaction would exclude someone. The interview schedule was developed with autistic women, clinicians, and researchers. It covered the path to diagnosis, relationships, sensory experience, mental health, and participants’ views of gender differences.

The researchers used Framework Analysis. They coded the transcripts, compared cases, and organised the material into four main themes and 19 subthemes. The design goes deep into meaning and context; it does not measure prevalence or compare groups.

The four main themes were being told or treated as though they could not be autistic, trying to fit in, moving from passivity toward assertiveness, and forming an identity as an autistic woman.

Twelve participants described being given other labels before autism, and 12 described professional stereotypes that got in the way. Six said being quiet at school helped them go unnoticed. Several remembered looking composed in class and falling apart at home, or being treated as lazy or rude when they misunderstood an expectation.

Camouflaging was common but not universal. Eight accounts were coded under “wearing a mask,” six under learning social behaviour from television, books, or magazines, and five under social mimicry. Participants described both deliberate strategies and copying they did not initially recognise as copying. Five accounts connected masking with costs such as exhaustion, confusion about identity, or needs remaining invisible.

The interviews also contained serious accounts of vulnerability. Eight participants described entrapment in abusive relationships or risky situations, and nine disclosed sexual abuse. Those numbers describe this particular group. The study was not designed to estimate population risk or show that autism, masking, or late diagnosis caused the abuse.

Not how common these experiences are. Recruitment deliberately sought women who had been diagnosed late. The study cannot tell us what proportion of autistic women mask, are missed, or share any of the four themes.

Not a comparison between genders. There was no group of autistic men, non-binary people, or women recognised early. The paper generated hypotheses about gendered recognition; it did not test whether the traits were unique to women.

Not the full autistic population. Participants were young adults, mostly in work or education, and all could complete an interview. The authors noted that people with intellectual disability, people with higher support needs, older women, and those whose autism was recognised early might have different experiences.

Not a causal account. Participants looked back on years before diagnosis. Their explanations are important evidence about experience, but the design cannot isolate why one person was missed or show that masking caused a delayed diagnosis.

The paper’s durable contribution is not a list of “female signs.” It is the shift in viewpoint. A person can be working hard to look socially fluent while still experiencing confusion, sensory load, or a need for support. Quiet behaviour can make strain less visible. Professionals can also bring a narrow prototype to the room and mistake a mismatch with that prototype for absence of autism.

The most responsible use of this study is to widen inquiry, not narrow it again. Ask how someone learned the social rules, what the performance costs, what happens after the public-facing effort is over, and which needs have stayed hidden. Do not turn the 14 participants into a new stereotype that every autistic woman must match.

The language has aged, too. The paper repeatedly uses functioning labels and frames some differences as deficits. We can retain the design and findings without carrying that language forward.

This study represents 14 UK women, aged 22 to 30, who had a formal autism diagnosis, received it late, and volunteered for research about that experience. It is especially informative about people who could conceal enough difficulty to avoid recognition and later explain that process in an interview.

It does not represent all autistic women, all late-diagnosed people, non-binary or trans autistic people as a group, people with intellectual disability, or people who communicate in other ways. Its themes may resonate beyond the sample. Resonance is not the same as prevalence.

Studies should compare people recognised early and late, include varied genders and support needs, and follow people over time rather than reconstructing the recognition process years later. Measures of camouflaging should sit beside interviews so researchers can test whether visible social behaviour, self-reported effort, and diagnostic decisions actually move together.

We would also want assessment research to test the systems, not only the people. When two clinicians see the same history, which details change their decisions? Does training reduce gendered referral bias? And do broader assessment practices improve useful support rather than merely changing who receives a label?