Two major autism publications, in just one week
Within days of each other, two significant publications on autism have challenged how we understand it. The first, published in Nature, demonstrates that people diagnosed with autism after age ten have genetically different profiles from those diagnosed in childhood. The second, from Aimee Grant in The Conversation, argues the entire concept of an “autism spectrum” has outlived its usefulness.
Together, they dismantle the comfortable fiction that autism exists on a single line from “mild” to “severe,” with late-diagnosed people simply occupying the milder end.
The Nature study compared genetic profiles of thousands of autistic people in Denmark and the United States.
The finding: late-diagnosed individuals are genetically more similar to people with ADHD than to those diagnosed with autism in childhood. They also show higher risk for depression.
This isn’t about delayed recognition of the same condition. It’s evidence that “autism” may describe multiple distinct phenomena with different underlying biology.
Late diagnosis reveals different genetic profiles
Check out our NHS assessment data interactive map.
For years, the standard explanation for late diagnosis went like this: these individuals have autism, but their symptoms are mild or subtle enough — or they were functioning or simply overlooked enough — to go unnoticed, unacknowledged, and untreated during childhood. Only when life demands increase — social complexity, executive function requirements, masking exhaustion — do their difficulties become apparent enough for diagnosis.
The Nature research suggests this explanation is incomplete. If late-diagnosed people were simply “milder cases” of the same condition, their genetic profiles should resemble childhood-diagnosed individuals.
They do not.
Thomas Bourgeron, study co-author at France’s Pasteur Institute, stated directly: “People diagnosed with autism later in life are genetically more similar to those with attention deficit hyperactivity disorder.”
This has implications for anyone diagnosed after childhood — myself included, diagnosed at 21. The traits, challenges, and experiences that led to late diagnosis may not be diluted versions of childhood autism. They may be fundamentally different neurological configurations that happen to meet current diagnostic criteria for “autism spectrum disorder.”
The genetic overlap with ADHD is particularly relevant for those of us who are AuDHD. Rather than having two separate conditions that happen to co-occur, the biology might be more integrated than current diagnostic categories acknowledge.
So, we just about fit into a pre-existing framework (which is how science works), and that’s that. Diagnosed. AuDHD. Bosh.
The linear spectrum was always misleading
The “autism spectrum” metaphor, coined by psychiatrist Lorna Wing in the 1980s, was revolutionary for its time. It expanded understanding beyond rare, narrowly-defined cases to recognise a wide range of autistic experiences.
But as Aimee Grant argues in The Conversation, the metaphor has calcified into something misleading. When people hear “spectrum,” they picture a straight line — like colours arranged from red to violet. Applied to autism, this creates a ranking system from “more autistic” to “less autistic.”
That’s not how autism works. Autism comprises multiple traits appearing in unique combinations: some people rely heavily on routine, others on stimming, others on monotropic attention to specific interests. Physical traits like hypermobility often accompany autism. The point is that no single line can capture this multidimensional reality.
The American Psychiatric Association’s attempt to impose structure through three “levels” (requiring support, substantial support, very substantial support) only creates new problems. Research shows these levels are vague and inconsistently applied. They don’t reflect real-world experiences.
More importantly, they’re static. Support needs fluctuate based on life circumstances, autistic burnout, life stages like menopause, and whether needs have been met or ignored over time. A “level” assigned at one point becomes meaningless as circumstances change.
Support needs aren't static measurements
The recent proposal of “profound autism” by The Lancet Commission as a category for autistic people with learning disabilities or high support needs illustrates the ongoing urge to create hierarchies. But as critics note, the phrase tells us nothing useful about specific challenges or required support.
What it does do is divide autistic people into implicit value categories. Those with lower support needs become the “acceptable” autistic people — employable, communicative, less burdensome. Those with higher support needs get segregated into “profound” territory, linguistically separated from the broader autistic population, and understood as those who need to be put into the different schools and systems of support.
Grant warns this categorisation “can slip into judgments about their value to society. In the most extreme form, such hierarchies risk dehumanising those with higher support needs.”
The warning isn’t theoretical. Hans Asperger, whose name is attached to “Asperger’s syndrome”, was responsible for the Nazi genocide of autistic people — all those who Hans described as “autistic psychopaths” — with higher support needs. I bet you didn’t expect to read that, right? Crazy.
The separation of autistic people into “useful” and “burdensome” categories has historical precedent with catastrophic outcomes. Obviously.
Grant notes that current US political direction is “very worrying to many autistic people.” RFK Jr.’s framing of autism as an “epidemic to confront” — combined with demonstrably false claims about Tylenol (paracetamol) causing autism — reveals how quickly medical language can slide into implied threat and “required” response.
When authorities frame autism as a crisis requiring elimination rather than accommodation, the spectrum metaphor becomes dangerous. It creates tiers of humanity where some autistic people might be tolerated while others are targeted for prevention or cure.
Why language matters beyond politeness and progressiveness
People often use “on the spectrum” as euphemism, avoiding the word “autistic” because they consider it negative. This well-meaning avoidance reinforces the stigma it claims to circumvent.
Many autistic adults prefer direct language: autism, autistic. Not because the clinical terms are particularly good — they’re not, and again, are just a box — but because euphemism implies there’s something shameful to obscure.
The Nature study’s finding that late-diagnosed autism may be genetically distinct from childhood-diagnosed autism doesn’t resolve the language problem. It complicates it. If autism describes multiple phenomena, what should those phenomena be called? Should late-diagnosed people continue identifying as autistic when their genetic profiles differ from childhood-diagnosed populations?
These aren’t just semantic questions. Language shapes access to services, research funding, community identity, and legal protections. It determines who gets accommodated and who gets pathologized. Whether we like it or not, what we call things — and often most significantly, ourselves — is important.
The convergence of these two pieces of research in neurodiversity — one demonstrating genetic differences in late diagnosis, another arguing against the spectrum metaphor — points toward a necessary reckoning with how autism is conceptualised and discussed.
Moving away from linear spectrum thinking means recognising autism as multidimensional, context-dependent, and potentially comprising several distinct neurological profiles currently grouped under one umbrella term.
The alternative to the spectrum isn’t a return to narrow, exclusionary definitions. It’s acceptance that human neurology is too complex for tidy hierarchies, and that attempts to rank autistic people from mild to severe serve social sorting rather than understanding or support.
Update — May 2026
Seven months on, the argument has gathered the corroborating evidence the original Nature finding hinted at. The premise that late-diagnosed autism is not a milder version of childhood-diagnosed autism but a genetically and developmentally distinct trajectory has held up.
The Cambridge findings examined in genetic timing — when autism becomes visible is written in DNA extended the picture: timing of emergence is itself heritable, with two clusters of genetic influence — one tied to early-diagnosed autism with greater developmental load, and one tied to later-diagnosed autism with stronger links to higher educational achievement and subtle social differences. Different routes, different rhythms, same diagnostic label.
The Fonagy Review’s interim report then tested whether that picture would survive contact with policy. The Directory’s response, the three-level neurodivergent severity framework and what the overdiagnosis debate misses, made the structural point: epidemiological prevalence has stayed stable while administrative diagnosis has risen, because the population of late-emerging Level 1 and Level 2 individuals is finally being counted.
And on the spectrum-as-hierarchy concern, Spanish research on twice-exceptional adults camouflaging at autism-equivalent levels confirms what the linear model obscures — visible competence is not reduced need.
The spectrum metaphor was always too neat. The biology refused to cooperate.
Citations
Nature — Polygenic and developmental profiles of autism differ by age at diagnosis
The Conversation — Why it’s time to rethink the notion of an autism ‘spectrum’ (Aimee Grant)
The Lancet Commission — the future of care and clinical research in autism
