Did we miss the girls?

Earlier this year, a high-profile study published in the British Medical Journal reported that the male-to-female ratio of autism diagnoses has been steadily narrowing. Registry data across a 35-year period showed an increasing number of females diagnosed with autism, such that the male to female ratio would be equal by the age of 20 (Fyfe et al, 2026).

The research was widely interpreted as confirmation of something many people already believed: that girls had always been autistic in similar numbers to boys, but had simply been overlooked. Media commentary suggested we were finally recognising “the missing girls”.

This is an appealing explanation.

But it rests on an assumption: that the autism being diagnosed today is the same construct as that being diagnosed 40 years ago.

That assumption is rarely stated explicitly. Instead, it remains in the background, shaping how changes in the sex ratio are interpreted. However the diagnostic construct we now call Autism Spectrum Disorder (ASD) or ‘autism’ (a semantic issue I will leave for another time) has changed repeatedly over the past four decades. If the construct itself has changed, then comparing sex ratios across time becomes more complicated than it first appears.

Before we ask whether girls were missed, there are two questions to consider;-

· Were individuals diagnosed at different ages eligible for the same diagnostic criteria during childhood?

· And does a changing sex ratio necessarily indicate improved ascertainment, or could it also reflect changes in who now meets diagnostic criteria?

Autism first entered the DSM in 1980 as Infantile Autism. Autistic Disorder followed in 1987, Asperger’s Syndrome in 1994, and Autism Spectrum Disorder (ASD) in 2013. The name changes were not simply window dressing.

Each revision significantly altered who could receive a diagnosis, and what symptoms or presentation met the criteria. Importantly, ASD broadened the construct to a range of traits on a dimensional spectrum, allowed diagnosis across the lifespan, and introduced the possibility that neurodevelopmental features may only become apparent when social demands exceed capacity, or may not have been observable due to masking.

The sequence of dates matters. They demonstrate that many teenagers and adults described today as “historically undiagnosed autistic people“ were not necessarily overlooked. Instead, the diagnostic category under which they were eventually diagnosed did not yet exist.

Consider a woman who is 50 years old today. During childhood she could have been diagnosed with Infantile Autism or, later, Autistic Disorder if she had shown marked developmental impairment. Asperger Syndrome was only introduced as she entered adulthood, while ASD did not become available until she was almost forty. If she is diagnosed with ASD today, it does not automatically follow that she was “missed” throughout childhood. The diagnostic construct under which she eventually qualified simply did not exist for much of her life.

Now consider a woman who is 26. During her childhood, clinicians could diagnose Autistic Disorder or Asperger Syndrome. At the age of 13, however, she became eligible to be assessed under the newly introduced ASD criteria. If she was diagnosed after that point, it may reflect not only increased recognition, but also the introduction of a different diagnostic construct with different eligibility criteria.

Concerns that girls might present differently were not unreasonable. Since the introduction of Infantile Autism in 1980, a 4:1 ratio of boys to girls had been observed, although that ratio dropped when the children had concurrent intellectual impairment (Fombonne 2009).

During the early 1990s researchers began proposing hypotheses to explain why boys under 4 were diagnosed much more often than girls under 4. Perhaps boys were genetically disposed toward the condition. Perhaps girls expressed autism differently. Perhaps they possessed some form of biological protection. Perhaps clinicians were overlooking them.

These were hypotheses. Several were plausible. The question is how one particular hypothesis gradually came to organise the way later findings were interpreted.

When Asperger Syndrome entered the DSM in 1994, diagnoses still favoured males, although more females were represented than in Autistic Disorder.

Shortly after the introduction of ASD, a number of influential papers appeared in scientific journals, setting the scene for future research to explore gender differences in ‘autism’. Researchers suggested there may be a unique female phenotype, and that girls and women may present slightly differently, which may account for the reason they were under represented in Autistic Disorder and Asperger Syndrome. In particular, it was proposed that girls had better social skills, might mask their neurodevelopmental features, and had different interests, which may contribute to the lower rate of diagnosis.

The progression of the arguments proposed went a little like this:-

1. Boys are diagnosed more often than girls. (Observation)

2. Girls may have protective biological factors. (Testable hypothesis)

3. Girls may camouflage or mask autistic traits. (Testable hypothesis)

4. Girls may therefore present differently. (Inference)

5. Clinicians may therefore have overlooked autistic girls. (Inference)

6. Large numbers of undiagnosed autistic women may therefore exist. (Further inference)

7. Later-diagnosed women become evidence that the original hypothesis was correct. (Potentially circular argument unless independently demonstrated)

None of these steps is unreasonable on its own. But each step moves a little further from direct observation. Over time, a chain of plausible inferences can gradually come to feel like an established empirical fact. Particularly when they are repeated often.

During the years following the introduction of ASD, the discussion shifted beyond scientific hypotheses toward a distinctive narrative of recognition. Across high-profile journals, review articles and professional commentaries, a recurring vocabulary emerged.

• Many females who, if skilfully assessed, would meet full diagnostic criteria, never receive a diagnosis.

• A lost generation of autistic adults.

• Women were finally being recognised.

• Women mask their autistic traits.

• Girls’ better social skills prevented appropriate healthcare.

Each of these statements may be a plausible interpretation of the available evidence. Collectively, however, they do something more.

They invite a particular reading of history: that many women diagnosed under contemporary ASD criteria would also have met the diagnostic criteria of previous decades, had clinicians (or teachers) simply recognised them.

This is a powerful historical claim. However, it remains an inference rather than a directly established historical fact.

As these ideas moved beyond scientific journals into newspapers, documentaries, advocacy organisations and social media, the language often became more definitive. What had begun as a series of hypotheses increasingly came to be presented as an historical account of what had happened. Over time, a biological observation started to become framed as a gender equality issue.

There is an important difference between saying:

Some women are now being diagnosed under broader ASD criteria.

and saying:

These women were autistic all along but were historically missed.

The first is an observation about contemporary diagnostic practice.

The second is a historical interpretation of why those diagnoses occur.

Those two propositions are often treated as interchangeable, but they are not.

Once this narrative became culturally familiar, diagnosis came to offer something more than a clinical explanation. It also offered a way of understanding one’s life.

“You were missed.”

“Things might have been different.”

“Your struggles resulted from decades of being undiagnosed.”

For many people this is profoundly validating. The emotions are real. But the story attached to those emotions remains an interpretation of history, rather than historical evidence in itself.

Because the criteria under which many women were later diagnosed was not available for much of their lives.

In Norway, Prosserud and colleagues examined the male to female rate of diagnosis from for Autistic Disorder, Asperger’s Syndrome and Intellectual disability over a 20-year period, concluding in 2015.

The results were striking.

Across more than four decades, the male-to-female ratio for Autistic Disorder hardly changed at all. It remained remarkably stable at around 3.6 boys for every girl diagnosed.

Asperger’s Syndrome told a different story. The MFR in this group was 2-3:1. More females appeared in this group over time, although boys still outnumbered girls.

In other words, long before Autism Spectrum Disorder existed, male-to-female ratios already differed depending on which developmental presentation you were looking at.

These findings are quite similar to data from the United Kingdom.

The recent Swedish study by Fyfe and colleagues showed that the MFR of children diagnosed under the age of 4 remained remarkably stable across 40 years – at 4:1. The change didn’t occur in the preschool years. It didn’t even occur before age ten. The shift happened primarily during adolescence, with the largest increase in female diagnoses occurring between about 15 and 19 years of age, after the introduction of ASD.

These observations do not prove that girls were not missed. However there is a pattern here that warrants examination.

If improvements in assessment and greater awareness are now enabling clinicians to identify girls who were previously overlooked, why has the male-to-female ratio remained so stable among childen under 10?

What is interesting, is that prevalence data seem to emit the same ‘signal’ that was present before the ASD classification emerged. If that is the case, examining different sub-groups as a whole will obscure meaningful differences in male to female ratios – making it harder to interpret trends in prevalence, genetics, and outcomes.

Some possible research questions:

• Does the stable MFR among preschool children indicate a genuine biological sex difference in the early developmental phenotype?

• Does the recent convergence in MFR reflect improved recognition of the same population, or expansion of the diagnostic construct to include a different developmental population?

• How similar are females diagnosed under contemporary ASD criteria to females who would previously have met criteria for Autistic Disorder or Asperger Syndrome?

Another recent prevalence study points towards a significant change in the autism construct. Salkic et al (2026) found the co-occurrence of Autism and Intellectual disability remained stable across 20 years, in the context of an 800% rise in prevalence in those without cognitive impairment. The average age of diagnosis doubled, from 8 to 14, and the study cited the finding that autism was now diagnosed with 50% less symptoms. These authors were remarkably candid, noting that the autism phenotype of today captured a substantially broader and more cognitively heterogenous population than 20 years ago. Further, they suggested that prevalence increases cannot simply be interpreted as the discovery of previously missed neurodevelopmental pathology, but may be ‘substantially’ due to administrative and secular changes.

Science advances by proposing explanations and then testing them against competing alternatives. But epistemology—the lens through which we understand the world—also shapes the questions we ask, the methods we choose, and the explanatory models we use to interpret our findings.

The proposition that girls were historically overlooked was an important and entirely reasonable hypothesis. Over time, however, that hypothesis became the dominant interpretive framework through which later findings came to be understood.

Across 45 years of research (Fombonne, Prosserud, Salkic and Fyfe), one finding has remained remarkably consistent: the early developmental phenotype continues to show a stable male-to-female ratio, while convergence occurs predominantly among later diagnoses. The ratio also becomes less male-biased when autism occurs alongside intellectual impairment.

Meanwhile, we are witnessing a rapid increase in later-diagnosed females.

Before concluding that we are finally recognising the girls who were always there, we first need to ask whether they were ever eligible for the diagnosis we are retrospectively applying.

Until then, “Did we miss the girls?” remains an important scientific question—not an established historical fact.

Fombonne E. Epidemiology of pervasive developmental disorders. Pediatr Res. 2009;65(6):591-598.

Fyfe C, Winnell B, Dougherty M, et al. (2026) Time trends in the male to female ratio for autism incidence: population based, prospectively collected, birth cohort study. BMJ. 2026;392:e084164.

Prosserud MB, Skretting Solberg B, Engeland A, Haavik J, Klungsøyr K. (2022) Male to female ratios in autism spectrum disorders by age, intellectual disability and attention-deficit/hyperactivity disorder. Acta Psychiatr Scand. 2021;144:635-646.

Salkić L, Tideman M, Martini M, et al. (2026) The proportion of intellectual disability in autism spectrum disorder over two decades. Psychiatry Research, Volume 363,2026.