What If the Autism Spectrum Is Wrong?

Three words have come to define how most people think about autism: 'on the spectrum.' The phrase carries with it a particular mental image: a single, continuous line stretching from one kind of autistic experience to another, with every person somewhere along the range.

The problem, as a growing body of research is now making clear, is that this image may be fundamentally wrong.

A January 2026 investigation in New Scientist put the question directly: what if the idea of the autism spectrum is completely wrong? The article pointed to something that many researchers had been quietly grappling with for years. Autism's extraordinary diversity, the fact that some autistic people do not speak at all while others are hyperverbal, that some are acutely sensitive to light and noise while others show the opposite pattern, that some have rigid routines and repetitive movements while others spend hours absorbed in intensely specific interests, has always been difficult to explain if everyone shares the same underlying condition to a greater or lesser degree.

Several recent studies are now suggesting they do not.

Four groups, not one spectrum

In July 2025, a team of researchers at Princeton University and the Flatiron Institute published findings in Nature Genetics that have reframed the conversation. Analysing data from 5,392 autistic children aged 4 to 18 through SPARK, the largest autism research database ever assembled, they applied machine learning across more than 230 traits covering genetics, development, and behaviour. What they found was not a spectrum. It was four distinct groups.

The genetic differences between these groups showed, in the words of senior author Professor Olga Troyanskaya of Princeton University, 'very little biological overlap.' These were not points on a continuum. They were distinct biological profiles.

Co-lead author Aviya Litman noted that despite autism being 'between 60 and 80 per cent heritable,' genetic causes remain unclear for approximately 80 per cent of autistic individuals tested. The study was validated against the Simons Simplex Collection, a separate dataset, confirming the subtypes reflected real biological differences.

The pioneer who helped build the spectrum is questioning it

Around the same time this research was gaining attention, another challenge to the spectrum came from an unexpected source. Dame Uta Frith, the developmental psychologist and neuroscientist who helped establish modern understanding of autism, gave an interview to the Times Educational Supplement in March 2026 in which she said she no longer believes autism is a spectrum.

Frith is not a peripheral figure. Nearly 60 years ago, she identified autism as a lifelong neurodevelopmental condition rather than the result of emotionally distant parenting, overturning the damaging 'refrigerator mother' theory that had caused enormous harm to families. Her 1989 book Autism: Explaining the Enigma became a foundational text, and her work on theory of mind shaped clinical practice for a generation.

Her concern now is that the spectrum has expanded so far beyond its original, precisely defined boundaries that it has become scientifically unreliable. 'The spectrum has gone on being more and more accommodating,' she said. 'And I think now it has come to its collapse.'

Frith proposes that the current diagnostic population is better understood as two distinct groups. The first: children identified in early childhood, often with intellectual disability and significant developmental differences. The second: a group diagnosed much later, predominantly adolescents and young women without intellectual impairment, who have strong verbal and non-verbal communication skills but experience high social anxiety and hypersensitivity. She argues that grouping these two populations under a single diagnostic label creates noise in research data. 'This makes the data we obtain from large groups very noisy,' she told The Times.

The counterarguments are serious

Frith's views have generated significant disagreement, including from researchers and autistic people with direct expertise in the field.

Professor Sue Fletcher-Watson, a developmental psychologist at the University of Edinburgh with approximately 20 years of autism research experience, published a direct response in April 2026. Her core argument is that autism cannot be understood independently of autistic people's own experiences, and that treating self-reported accounts as scientifically unreliable creates its own methodological problem.

Fletcher-Watson also points to where some of the most significant recent advances in autism understanding have actually come from. The double empathy problem, the concept of monotropism, and detailed research on masking all emerged primarily from autistic self-report, not neurobiological investigation. Narrowing the diagnostic criteria risks excluding the very people whose accounts have driven progress.

She also notes that for many autistic people, the diagnosis serves purposes beyond clinical classification. It provides community, shared identity, and the basis for political advocacy and rights recognition. Any move to divide the diagnostic category carries social and practical consequences that researchers have a responsibility to consider.

Other researchers, while supportive of the subtype research, have urged caution. Fred Volkmar, a psychiatrist at Yale University, described the findings as encouraging evidence that the field needs 'more fine-grained approaches to diagnosis.' Catherine Lord, a psychologist at UCLA, said the subtype groups 'make sense and follow lots of findings by other researchers,' though she stopped short of calling them a definitive framework. Michael Lombardo cautioned that the real challenge lies in 'unravelling the biology that stems from those genes.'

What this means for families and educators

For families and educators, this debate is not purely academic. The language used to describe autism shapes how children are assessed, how schools allocate support, and how families navigate systems like the NDIS.

If autism is better understood as four or more distinct biological profiles rather than a single spectrum, the implications are significant. Support that works for one subtype may not be appropriate for another. The Princeton study specifically noted that some children 'seem very neurotypical until a bit later in childhood,' a finding with direct consequences for when and how assessments are conducted. Children who appear to be managing well in early primary school but who carry genetic variants that express later in development may be going unidentified, or receiving support designed for a different profile entirely.

At the same time, the concerns about narrowing the diagnostic category are equally practical. If a tighter definition of autism excludes late-diagnosed women and girls who have spent years masking significant differences, those people do not stop having support needs. They simply stop having a clear pathway to access them.

Where things stand

The Princeton study is one of the largest of its kind, and its publication in Nature Genetics carries significant weight. Uta Frith's willingness to publicly question a framework she helped construct deserves serious engagement rather than dismissal. The concerns raised by autistic advocates and researchers like Fletcher-Watson about whose experience is centred in these decisions are not peripheral to the science. They are part of it.

What is clear is that describing autism as a single, continuous spectrum has always been a simplification. The question the field is now grappling with is what to replace it with, and whether any replacement will serve the full range of people it is meant to describe.

References