Autism in Adults

What “spectrum” actually means

Most people picture the spectrum as a line. Mild at one end, severe at the other, everyone placed somewhere along it. That picture keeps a lot of adults from ever getting assessed. They look at the line, decide they would sit too far to the left to count, and stop reading.

It was never one line

The manual codes several things at once. DSM-5 folded four older diagnoses into one: autistic disorder, Asperger’s disorder, childhood disintegrative disorder and PDD-NOS. Then it attached specifiers that vary independently of each other. With or without intellectual impairment. With or without language impairment. Associated genetic or medical conditions. Co-occurring mental health conditions. Two people with the same diagnosis can differ on all four. You cannot draw that as a line, and the manual does not try.

If you were told you had Asperger’s before 2013, that diagnosis did not expire. The manual says so directly: a well-established DSM-IV diagnosis of autistic disorder, Asperger’s or PDD-NOS should now be recorded as autism spectrum disorder.

American Psychiatric Association, Highlights of Changes from DSM-IV-TR to DSM-5 (2013); DSM-5-TR criteria for 299.00.

What the support levels mean

DSM-5 does define three levels: requiring support, requiring substantial support, requiring very substantial support. They get quoted as a measure of how autistic someone is. That is not what they measure. They describe how much support a person needs. The working group chose that framing deliberately, because an autistic person may be doing well precisely because support is already in place.

A level is also a snapshot. The APA’s own rating tool asks clinicians to score severity as experienced over the past seven days, then to score it again at intervals to track change. Seven days is not a lifetime.

American Psychiatric Association, Clinician-Rated Severity of Autism Spectrum and Social Communication Disorders.

“High functioning” has a related problem. A study of more than two thousand autistic people found that IQ barely predicted how someone actually copes day to day. Among those without an intellectual disability, coping scores sat well below what their IQ would suggest, at every age. The authors argued the term should be dropped from research and clinical practice.

Alvares et al. (2020), Autism 24(1), 221–232. The sample was aged 1–18; the finding is about the label rather than about adults specifically.

In fairness to the levels, one clinical study of 158 autistic people found that psychologist-assigned levels did track adaptive functioning in a graded way. What nobody appears to have tested is whether an adult’s level holds steady from one assessment to the next.

The NHS does not use levels

Levels 1, 2 and 3 belong to DSM-5, an American manual. The NHS diagnoses against ICD-11, which has no equivalent. ICD-11 subdivides autism by whether a disorder of intellectual development is present, crossed with functional language. So if you are in the UK and trying to work out your level, your health service may not have one for you.

WHO ICD-11, 6A02. UK clinical guidance: NICE CG142, Autism spectrum disorder in adults.

How Autism presents in adults

How the wiring actually differs

Confident claims outrun the evidence here more than anywhere else in autism writing. Autistic and non-autistic brains do differ measurably. The differences are smaller and more specific than popular accounts suggest, and the best-known explanations are still arguments rather than findings.

Attention that runs deep

Monotropism says autistic attention pools into a few interests at a time, and that most of the diagnostic criteria follow from that one difference: the social difficulty, the cost of being interrupted, the depth of interest. Three autistic researchers developed it. Among autistic adults it is often the description that lands hardest.

As science it is early. The founding paper is theoretical rather than experimental. The questionnaire built to measure it is recent, and the National Autistic Society is clear that it is not a diagnostic tool. We use monotropism here because a great many autistic people recognise it in their own attention, not because it has been proven.

Murray, Lesser & Lawson (2005), Autism 9(2), 139–156. Monotropism Questionnaire: Garau et al., preprint.

Global processing takes longer

You will read a lot about autistic attention to detail. On puzzle tasks — spotting a shape hidden inside a busy picture, rebuilding a pattern from blocks — autistic people do score a little higher on average. Small gaps between group averages, not a different kind of vision. The studies also disagree with each other more than anyone can explain.

The largest review, pooling 56 studies, found something different. Autistic people are no better at seeing the parts and no worse at seeing the whole. They are slower to assemble the parts into a whole. That fits the experience of registering every separate thing in a room before the room turns into a scene. It explains why busy places cost you more, without claiming you see detail other people cannot.

Van der Hallen et al. (2015), Psychological Bulletin 141(3), 549–573; Muth, Hönekopp & Falter (2014), JADD 44(12).

Connectivity, and the half of the sentence that holds up

The line you will see repeated is that autistic brains are heavily wired within regions and lightly wired between them. Reviewing 52 brain-imaging studies, researchers found decent support for the second half: less long-range wiring between distant regions. On the within-region half they concluded nobody yet knows. The studies used such different methods that the reviewers could not even pool them properly.

O’Reilly, Lewis & Elsabbagh (2017), PLOS ONE 12(5), e0175870.

Prediction

One influential idea is that autistic brains build and update predictions differently, which would explain why unexpected change costs so much. Reviewing 47 studies, researchers did find differences, strongest when the thing that signals what is coming is faint or inconsistent. They also said the idea needs defining far more tightly before it can be tested properly. A later study found autistic adults picked up hidden patterns exactly as well as anyone else. Doing something differently does not have to mean doing it worse.

Cannon et al. (2021), Autism Research 14(4); Pesthy et al. (2023), Scientific Reports 13, 11873.

Intelligence, and two opposite stereotypes

Autism and intelligence have been tangled together badly in both directions. First an assumption that most autistic people have an intellectual disability. More recently an assumption that autistic people are all quietly brilliant. Neither matches the figures. The second one costs autistic people support they need, because they present as capable.

Where the 70% figure went

Older studies reported that as many as 70% of autistic people also had an intellectual disability. That number is about forty years out of date, and three sets of diagnostic criteria behind. Newer work puts it at closer to a third. Here is where autistic IQ actually falls.

Sweden keeps health records for its whole population, so you can watch the change happen. Over twenty years, the share of autism diagnoses that came with an intellectual disability fell to almost nothing. Autistic people did not change in those twenty years. The group being counted did, as the criteria widened and adults started being recognised at all.

Zeidan et al. (2022), Autism Research 15(5); Shaw et al. (2025), MMWR Surveillance Summaries 74(SS-2); Salkić et al. (2026), Psychiatry Research 363. The ~70% figure as reported by Thurm et al. (2019), Frontiers in Psychiatry 10, 526.

And no, not all quietly brilliant either

In one large study of autistic children, only three in a hundred had an IQ above 115. No research shows autistic people as a group scoring unusually high. You will also see it said that most autistic people have one standout skill. That comes from parents and clinicians reporting anything a person is notably good at. It is a much wider net than savant ability, which is rare.

Charman et al. (2011), Psychological Medicine 41(3); Meilleur, Jelenic & Mottron (2015), JADD 45.

The measurement problem

What does hold up is that the usual tests may measure autistic ability badly. Clinics normally use the Wechsler scales. Give autistic people Raven’s Matrices instead — a visual puzzle test, no words involved — and their scores jump. Non-autistic people barely move between the two. These are small studies from one research group, and nobody has repeated them at scale, so hold them loosely. But if you have ever been tested and felt the number did not describe you, there is published reason for that.

Dawson et al. (2007), Psychological Science 18(8); Soulières et al. (2011), PLOS ONE 6(9), e25372.

Twice-exceptional

Twice-exceptional means gifted and disabled at once, most often alongside autism, ADHD or dyslexia. The mechanism is the part that matters here. The two cancel each other out on the measures schools use. Ability compensates well enough that no difficulty shows. The difficulty suppresses performance enough that no giftedness shows. The child comes out looking average, with a high ceiling and a real support need, neither of them visible. A lot of adults who say they coasted and nobody noticed anything are describing that.

Treat prevalence claims here with suspicion. The numbers in circulation are poorly founded, and the one formal modelling exercise concluded that twice-exceptionality is rarer than the literature implies.

National Association for Gifted Children (2009); Cheek et al. (2023), Exceptional Children 90(1).

A caveat on nearly all of this

Someone went through 301 autism studies from a single year. Almost none of the people in them had an intellectual disability, at a time when around half of autistic people did. Four in five of those studies were skewed this way, and many did not report ability at all. So almost every piece of research on this page describes autistic people without an intellectual disability. It probably describes you. The people it leaves out are the ones who need support most.

Russell et al. (2019), Molecular Autism 10(1), 9.

Empathy, emotion, and the double empathy problem

The claim that autistic people lack empathy has done more damage than any other single idea about autism. It is also wrong in a specific way, and the specifics are more use than a flat denial.

Which part is actually different

Empathy is not one thing. Noticing that someone is having an emotion, working out which emotion it is, feeling something in response, and responding in a way others recognise are four separate steps. The research puts the difference in the first two and the last. Affective empathy, actually feeling it, is the only component unique to empathy, and it is not different in autism. Two senior researchers in the field put it this way: autism is associated with atypical theory of mind but not empathy, while alexithymia is associated with atypical empathy but not theory of mind.

Fletcher-Watson & Bird (2020), Autism 24(1), 3–6.

Alexithymia: common, not universal

Alexithymia is difficulty naming and describing your own emotions. Pooling fifteen studies, it shows up in about half of autistic people and about one in twenty non-autistic people. The researchers’ own phrase is “common, rather than universal”.

The argument that follows is that emotional difficulty in autism tracks alexithymia rather than autism itself. Alexithymia scores, not autism severity, predict how accurately someone reads emotion in a voice. In one imaging study, once alexithymia was accounted for, the autism-related differences in empathic brain activity disappeared.

That has practical consequences. If naming what you feel is hard, it is a separate and describable thing with its own strategies, and roughly half of autistic adults do not have it. It also rules out the opposite stereotype. Autistic people are not, as a group, hyper-empathic either.

Kinnaird, Stewart & Tchanturia (2019), European Psychiatry 55, 80–89; Bird & Cook (2013), Translational Psychiatry 3(7), e285.

The double empathy problem

Damian Milton proposed that the breakdown between autistic and non-autistic people runs both ways: a mismatch between two kinds of mind, not a deficit sitting in one of them. A study of 72 adults tested it by passing a story down chains of eight people. All-autistic chains held onto the story about as well as all-non-autistic chains. Mixed chains lost it significantly faster, and rapport ratings dropped in those same chains.

That is one study across nine chains, and there is at least one recent critical review of this literature we have not read. The finding still matters, because it reframes something most late-diagnosed adults recognise. You are not uniformly bad at communicating. You are much better at it with some people than others, and who those people are is not random.

Milton (2012), Disability & Society 27(6); Crompton et al. (2020), Autism 24(7), 1704–1712.

Why adults are missed

What this might look like for you

What helps

Common misconceptions

If you are finding this out late

One study of autistic university students found that learning you were autistic earlier was associated with better quality of life. It was widely repeated. A larger pre-registered study of 300 autistic adults aged 18 to 68 then failed to replicate it. Age at learning did not uniquely predict quality of life or wellbeing. Trait level did.

So the fear that finding out at 35 or 50 means the good outcome has already gone is not supported by the evidence. What the qualitative work describes, in women diagnosed after 40, is two things at once. A shift from self-criticism to self-compassion, with a real sense of agency. And genuine grief at having arrived so late. Relief and loss in the same week is the ordinary shape of this.

Leung et al. (2024), Autism 28(2), failing to replicate Oredipe et al. (2023); Leedham et al. (2020), Autism 24(1), 135–146 (interviews with 11 women, UK).