Is autism a category or the far end of a spectrum?
Parents often ask me whether their child "has autism" or is "a bit autistic". Behind that question is an older one. Is autism a distinct group, like a broken bone? Or is it the far end of a range of traits everyone has to some degree, like height? A new review in JCPP Advances looks at the evidence from the best tool for the question. For autism, the answer is less settled than it first appears.
What the review looked at
Taxometric analysis is a set of statistical methods built for one job. It tests whether the pattern in the data points to a real boundary between two groups, or to a smooth continuum. It does not assume groups in advance.
Toffalini and colleagues, including Edmund Sonuga-Barke, searched for every peer-reviewed study that used these methods in neurodevelopmental conditions. They found 14. Together these held 85 separate analyses. Six studies looked at ADHD (attention deficit hyperactivity disorder), six at autism (one of which also looked at ADHD), two at language impairment and one at dyslexia.
What they found
The ADHD and language studies largely pointed to a continuum. This held across children, adolescents and adults for ADHD.
Autism looked different. Of the 29 autism analyses, 21 (72%) pointed to a category. All four studies that asked whether autism as a whole is a category said yes.
The problem with the autism studies
Here is the catch. All four of those autism studies used what the authors call artificial admixture. This means mixing people recruited in different ways into one sample. A typical example is a clinic group of diagnosed people added to a separate group of non-diagnosed people.
Across the whole review, analyses with admixture reached a categorical result 65% of the time. Analyses without it did so 15% of the time.
The authors then ran a simulation. They built data that was purely continuous, with no hidden category at all. They "diagnosed" the top 1% and drew 1,000 people from that population. Then they added a further 1,000 diagnosed cases, as a researcher short of cases might. Across 1,000 runs, the analysis came out in favour of a category every time. Without the added cases, it leaned towards a continuum, though often without a clear answer.
So a mixed sample can manufacture a category out of continuous data. The authors conclude that the evidence for a category is stronger for autism than for ADHD or language. But a firm conclusion is not possible while the main studies carry this bias.
A point from genetics
The authors make a point I recognise from my own work on the genetics of autism. Most autism liability looks polygenic, built from many common variants of small effect. A smaller group carry rare variants of large effect, often with wider developmental difficulties. That group could produce something that looks like a category in clinic samples. None of the reviewed studies tested this directly.
Why this matters for thresholds
If autism is a continuum, there is no natural line in the data. Any diagnostic threshold is then a clinical decision, not a discovered fact.
That does not make diagnosis meaningless. The authors make this point clearly for ADHD and language. A threshold can still be justified by the level of difficulty, the risks, and the likely benefit of support. It just should not be presented as a line drawn by nature.
This fits what I see in clinic. Children just below the threshold often have real needs. They are not a different kind of child from those just above it.
Take-home
- The best current evidence says ADHD and language impairment are continuous.
- Autism studies more often find a category, but nearly all of them used mixed samples that can produce false categorical results.
- Treat the diagnostic threshold as a practical decision about need and benefit, not a natural boundary.
- For the child just below the line, plan support on their difficulties, not on the label they missed.
Dr Odet Aszkenasy is a Consultant Community Paediatrician and the author of The Genetics of Autism: A Guide for Parents and Professionals.
Reference
- Toffalini E, Calderan M, Pagan R, Tobia V, Sonuga-Barke E. Categories or dimensions? A critical systematic review of taxometric evidence in autism, ADHD, and language disorders. JCPP Advances. Published online 8 October 2026: e70177. https://doi.org/10.1002/jcv2.70177