What if the idea of the autism spectrum is completely wrong?

“On the spectrum.” These three words have become synonymous with autism, yet behind them lies a common misunderstanding. The idea of “the spectrum” suggests that all autistic people share similar experiences and behave in similar ways – only to a greater or lesser extent. The reality couldn’t be further from the truth.

Some autistic people may not speak at all; others are hyperverbal and extremely fluent. Some are highly sensitive to bright lights and noise, or the opposite. And some have rigid routines and make repetitive movements like hand-flapping, while others are more flexible but spend a lot of time on “special interests” – anything from Tudor history to Rubik’s cubes.

Autism’s incredible diversity is something to celebrate. However, it has long presented an immense challenge to researchers trying to understand this seeming jumble of traits. Strides are now being made, as several recent studies have identified apparent groups within the catch-all term of autism that are also underpinned by patterns of genes and brain activity.

Researchers are exploring if and how these subtypes can be leveraged to help autistic people get better, more personalised support, and gain a great understanding of themselves. “There is now a more concrete basis for understanding where their experiences are coming from,” says neuroscientist Conor Liston at Weill Cornell Medicine in New York.

Yet this isn’t the first time that researchers have tried to separate autism into different guises and some advocates are wary of how these subtypes will play out in society. “You might feel like [subtyping] is value-neutral, but for someone else, it really isn’t,” says Amy Pearson, a psychologist at Durham University, UK.

Neurodiversity

Autism is a developmental condition that affects how people interact with others and the world around them. Broadly speaking, this means that autistic people often have difficulties with socialising, communicating and sensory sensitivity, and they may have restricted behaviours and interests. In numerous countries, autism is legally classed as a disability, which can help autistic people access support. However, many autistic people argue that it isn’t a disability, but instead a form of neurodivergence – and others are happy with both designations.

A growing awareness of neurodiversity has led to increased rates of autism diagnosis, with estimates now putting the number of diagnoses in the US at 1 in every 32 people. This includes many women and girls, who are often diagnosed later in life, as autism tends to present differently in them, especially when it comes to social motivations and behaviour. Traditionally, this hasn’t been accounted for.

In some ways, the idea of an autism spectrum, first coined by psychologists in 1979, fits this encompassing approach to diagnosis – and many autistic people still find the concept helpful. At the same time, the growing need to describe large variations in behaviour and experience has revealed the spectrum’s limitations.

Paul, a project manager from Maryland in his early 50s, is currently going through this diagnostic process. He struggles with interpersonal skills, such as “understanding what other people are feeling” if they don’t express themselves literally, he says. “It affects me at work, but it’s also helped me at work, because I ask a lot of questions so that I can understand something, and until I understand it, I don’t let it go.” Because of well-worn stereotypes about the spectrum and what autism is like, it never occurred to Paul that he might be autistic until his therapist suggested it. “I don’t think anybody fits all of this stuff,” he says.

Liston says autism is a large catch-all category that lumps together “people with probably many different kinds of molecular, cellular and brain circuit mechanisms”. In order to get a better handle on the underlying biology, we need to think about more precise ways of identifying it and embrace the condition’s heterogeneity, he says. This, in turn, could lead to earlier diagnosis and personalised support for autistic people. “Ultimately, that’s the goal,” says Adriana Di Martino at the Child Mind Institute in New York.

Searching for autism subtypes

So, in recent years, researchers have tried to demarcate autism subtypes by identifying clusters of people with similar sets of traits and symptoms, which may also have shared biological mechanisms. One early attempt was published in 2020 by developmental psychologist Mirko Uljarević at Stanford University in California and his colleagues. They asked the parents of 164 autistic children to rate their children’s social abilities and found five clusters that had distinct patterns of strengths and weaknesses across different social traits that didn’t map onto a simple line from more to less severe.

However, it became clear that this and similar studies could improve on their methods. Some studies relied heavily on parents’ reports about their autistic children, limiting their reliability. Moreover, it wasn’t clear in some research if these were true clusters or if the basic idea of a spectrum fit the data better. A 2020 review led by Di Martino concluded that there are probably “at least 2 to 4” distinct autism neurosubtypes, but that the studies were too small and relied on qualitative measures of autistic traits.

Since then, researchers like Di Martino have refined their methods, using larger sample sizes and identifying more granular behaviours and traits. They have also turned to brain imaging and genetic analysis to help match up behaviour with biological mechanisms. “We believe that is a more effective way to understand and characterise the features that are relevant for autism,” says Di Martino.

Now, what look to be genuine subtypes are appearing out of the fog. In a 2023 study, Liston and his colleagues analysed several existing datasets that comprised 432 autistic people whose brain activity had been measured and whose specific autistic traits had been identified. They reliably identified three distinct dimensions along which brain activity and behaviour were correlated in this group compared with a control group of neurotypical people – meaning people whose brains develop and work like most other people’s do.

One dimension related to intellectual functioning, especially verbal intelligence. The second was about social behaviour and relationships with other people, called “social affect”, and the third was linked to restricted interests and repetitive behaviours. Then, the team looked at how the autistic group scored on these three dimensions and found that their traits cluster into four subgroups.

Those in subgroup one had high verbal intelligence and strong connectivity in their language-processing centres, while it was the opposite for those in subgroup two. Likewise, while those in subgroup three had poor social affect but fewer restricted and repetitive behaviours, these traits were flipped in those in subgroup four. “Having identified those four subtypes, we can begin to ask questions about [why] they’re different,” says Liston.

Surprisingly, the team found that atypical connections in a given brain system didn’t lead to traits related to that system. “A lot of work to date has tended to assume that what is causing your symptoms, or what explains the severity of your symptoms, is also somehow abnormal,” he says. “And that is, in fact, not the case.” Some of the neurological changes may reflect one part of the brain compensating for problems elsewhere, says Liston. This detail would never show up in a study that lumped all autistic people together, he says, whereas the subgroup analysis revealed the underlying biology.

Genetic roots

Throwing genes into the mix offers further insights. Gene variants that are associated with autism often play a role in the connections that form between neurons, known as synapses. Liston’s team found that brain regions with altered circuitry in autistic people, compared with neurotypical people, also showed characteristic changes in gene expression. This implies that it should one day be possible to join the dots from genes to brain circuitry to behaviour, says Liston.

Another landmark subtyping study was published in July last year by geneticist Natalie Sauerwald at the Flatiron Institute in New York and her colleagues. They used a dataset from Simons Powering Autism Research, a research study that is led by the autistic community, which included 5392 autistic people – an order of magnitude more than previous studies. For each individual, the researchers examined 239 traits spanning seven categories: communication, restricted and repetitive behaviour, attention, disruptive behaviour, mood, developmental delay and self-injury. (It is worth noting that some autistic people feel that any greater incidence of self-harming, or other challenging behaviours among the autistic community is indicative of how they are treated by a world not built to support their needs.)

The researchers looked for patterns in the combinations of these traits and also found that they naturally fell into four subgroups, but these differed in several ways from the subgroups that Liston’s team found. Those in the first subgroup had a lot of difficulty with communication and restricted or repetitive behaviours, as well as disruptive behaviour, attention and anxiety, but no sign of developmental delay. Meanwhile, those in the second subgroup displayed developmental delay and a nuanced mix of other traits; those in the third subgroup had mild difficulties in all seven categories; and those in the final group had severe difficulties across the board.

Team member and geneticist Olga Troyanskaya at Princeton University and the Flatiron Institute says the researchers were surprised by how strongly the four groups came out of the data. “Every individual is unique, but there do seem to be these replicable groups.”

The idea that some subtypes might feature developmental delay was backed up by a study published in October last year, which looked at children diagnosed with autism between 5 and 17 years old. Researchers found evidence of two subgroups: those in subgroup one began experiencing social, emotional and behavioural difficulties early in life, while those in the other subgroup experienced upticks in difficulties in late childhood and early adolescence. These two subgroups were also linked to different sets of genetic variants, though there was some overlap between the two groups.

So far, it isn’t entirely clear how these and other subtyping studies fit together – or if there are two, three, four or more different subtypes. “I’ve sat down and tried to write down what each group is [in other studies] and how that fits with our groups,” says Sauerwald. Some of the categories her team identified are clearly distinct, but some seem to line up with those found by Liston and others. “We’re hopefully getting closer to the reality,” she says.

In one sense, the mismatch isn’t surprising, as these research teams have taken different approaches. Sauerwald and her colleagues concentrated on outward traits, whereas Liston and his colleagues focused more on connectivity within the brain. What’s more, the teams looked at different kinds of genetic variation: Sauerwald’s team examined changes in the genome itself, whereas Liston’s team looked at gene expression.

Switching subtypes

Confounding matters further, an autistic person might not stay in the same subtype all their life. “There’s a lot of clinical information pointing to changes over time and through development,” says Sauerwald. “As kids get older, sometimes they might switch.” In fact, one subtyping study published in 2024, which reassessed autistic people several years after an initial clustering assessment, found that nearly half of them changed subgroups within five years.

“I don’t think subtypes capture the multidimensionality of development,” says Pearson, who is herself autistic. Just like everyone else, autistic people change in a huge variety of ways over the course of their lives, so she says the subgroups can only ever be a crude approximation of behaviour and experience.

Indeed, it remains an open question whether truly distinct subtypes exist. These studies strongly suggest that some combinations of traits are more common than others, but it is still possible that every combination exists in somebody, somewhere. Because of these uncertainties, none of the subtyping researchers interviewed wants to see the subgroups they have identified used in healthcare clinics, at least not yet.

Still, with further advances, they hope their research can offer a framework that is helpful for the autistic community. Autism diagnoses already help many people make sense of themselves. Breaking down the broad autism diagnosis into subtypes could help autistic people understand each other’s varying experiences – and further validate their own, says Liston.

There has been greater celebration of autism and neurodivergence in recent years, enabled in part by the internet, and finer subgroups would be a natural extension of that, he says. Many autistic people are still told that they can’t really be autistic because they don’t have some trait or other, says Paul. With subtyping, “you would be able to say, ‘Well, I’m in this category’,” he says. Some might benefit from that, says Paul, although he doesn’t personally believe he needs it.

Troyanskaya, meanwhile, envisions doctors using subtypes to forewarn autistic people or their families of the specific challenges that they may encounter, perhaps years in advance. This would lead to “having the awareness to try to get support in place before the crisis, as opposed to after the crisis”, she says.

Another possibility that could one day emerge is targeted pharmacological treatments for specific adverse effects. This is a delicate topic because it may be conflated with the concept of a “cure” for autism, and that idea implies that being autistic is inherently bad. Many autistic people would say the condition is their strength, says Di Martino. Nevertheless, she argues that such treatments may be useful for some specific behaviours such as self-harm.

There is tentative evidence that autistic people in different subgroups respond differently to these treatments. For instance, one proposed drug suggested to improve social responsivity is the hormone oxytocin. So far, the results have been inconsistent. But a study published in 2024 that divided participants into two autism subtypes found that one group responded more strongly to oxytocin than the other. This may help explain oxytocin’s variable results, but won’t resolve the argument over whether pharmaceutical treatments are ever necessary, or if society needs to support autistic people more.

In a discussion in an online forum with New Scientist, a user called Neonatal RRT, an autistic hospital worker, wrote that more personalised approaches to healthcare are better, but that this also risks individuals slipping through the gaps if they don’t fit into finer categories. “People can be denied the care they need,” they wrote.

Asperger’s syndrome

Previous attempts at further categorisation haven’t worked out. Autism is currently an umbrella diagnosis used by doctors, but from 1994 to 2013, psychologists recognised a second, “milder” form of autism called Asperger’s syndrome, which was applied to people who lacked social skills and had restricted interests, but did acquire fairly typical language skills. Some autistic people still embrace the term “Asperger’s”, but many avoid it, either because of its namesake Hans Asperger’s links with the Nazi child euthanasia programme or because they disagree with the idea that autistic people can be divided into one group that needs less support and is perceived to be “high-functioning”, and another that is “low-functioning”. What’s more, some autistic people judged to be high-functioning sometimes found it harder to access treatment and support.

Anoushka Pattenden at the National Autistic Society in the UK is concerned that this new wave of subtyping research, while well meaning, could similarly backfire. “We fear that further categorising of autism is unhelpful and may lead to more stigma or discrimination,” she says. Pattenden, who is autistic, is glad that researchers have avoided labels such as high-functioning or low-functioning in these new subcategories, but says “you don’t have control over how that gets used, and what ends up happening with it”.

Sauerwald recognises these potential risks and says her team consulted with the autistic community when naming their subgroups. “We are constantly learning and doing our best to ensure that our work is beneficial to the communities involved rather than harmful, to the extent we can,” she says.

Ultimately, subtypes can only be beneficial if societies also become more empathetic towards autistic people, says Paul, which wasn’t his experience growing up. Pearson points out that many schools, universities and employers still don’t offer generic accommodations for autistic people, let alone personalised support.

“The first hoop to jump through with all of this is education,” writes Neonatal RRT, which may go some way to dismantling harmful stereotypes about autism. Instead of a uniform label of being “on the spectrum”, the autistic community, responding to the variety of their experiences, has alighted on another metaphor in recent years: the colour wheel. Every spoke of the wheel has a unique colour and represents an autistic trait, such as restricted interests and sensitivity to sensory stimuli, extending along the spoke to a different degree. In this way, the colour wheel, which contains many possible “plots”, underlines autistic individuality.

Sauerwald and other researchers hope that a respectful approach to subtyping can reveal autism’s underlying biology in a way that also brings this colour wheel, and the lived experiences it contains, into focus. What we choose to do with those subtypes – and how societies choose to treat neurodivergent people – is then up to all of us.