Can autism develop later in life?

What “early developmental period” requires

DSM-5-TR’s Criterion C asks that symptoms be present in the early developmental period. What it doesn’t ask for is that anyone noticed them, that they were written down, or that they caused visible problems at the time (American Psychiatric Association, 2022).

The criterion carries a parenthetical that matters as much as the requirement itself: symptoms may not become fully manifest until social demands exceed limited capacities, or may be masked by learned strategies in later life.

Read together, those two things say something specific. The traits have to be early. Their visible impact does not.

That distinction is what allows an adult with no childhood record, no early referral, and a school report full of praise to still meet the criteria. It’s also why a clinician assessing an adult spends so much of the appointment on childhood rather than on the present.

What “masked by learned strategies” looks like

It’s worth dwelling on that phrase, because it’s the diagnostic manual’s own wording rather than an advocacy gloss, and it describes something concrete.

The camouflaging research breaks it into two kinds of move. Masking is suppression: holding still instead of stimming, forcing eye contact, flattening a voice that wants to go up, editing out the thing you wanted to say. Compensation is substitution: running a mental script for small talk, copying a colleague’s phrasing, learning the rules of a social situation explicitly because they never arrived intuitively (Hull et al., 2017).

Masking and compensation can become practiced, and may make traits less visible to other people. Their presence doesn’t eliminate the developmental history, or the effort involved in producing the appearance—an effort that can be invisible to everyone, sometimes including the person making it.

This is part of why a childhood free of obvious difficulty proves less than people assume. Some autistic children are noticeably different. Others are quiet, rule-following, academically fine, and reading the room closely enough to pass, generating no concern because nothing about them is inconvenient to an adult.

Diagnosis usually happens in childhood, which sets the baseline

A systematic review and meta-analysis of studies published between 2012 and 2019 gathered age-at-diagnosis data across 40 countries. Pooling 35 studies covering 55 cohorts and 66,966 autistic people, it found a mean age at diagnosis of 60.48 months—five years old—with a range across cohorts running from 30.90 to 234.57 months (van ‘t Hof et al., 2021).

That’s the baseline against which adult diagnosis reads as unusual. Most recorded diagnoses happen in early childhood, and the services built around autism assume that timing. The wide upper end of the range is the part worth noticing, though: even within these samples, some cohorts were being diagnosed at close to twenty.

Two caveats belong with the figure. These studies largely sample children, so they describe when childhood diagnoses happen rather than the full lifetime distribution. And age at diagnosis says nothing about age at onset—it measures how long identification took.

Why a whole generation was missed

The clinical literature has a name for the adults now coming forward. A review in Lancet Psychiatry described the idea of a lost generation: people previously excluded from a diagnosis of classic autism who became recognisable as diagnostic criteria broadened, the spectrum concept arrived, and awareness rose (Lai and Baron-Cohen, 2015).

The same review is direct about why a first diagnosis in adulthood is hard to make. There may be no one available to give a developmental history. The person has usually acquired learnt or camouflaging strategies that obscure the traits. And co-occurring conditions are frequent, so the presentation arrives tangled with anxiety, depression, obsessive-compulsive difficulties, or a personality disorder diagnosis given years earlier. The review flags misdiagnosis as a particular risk in women.

None of those are reasons autism started late. They’re reasons it stayed unrecorded.

If it isn’t the autism that changed, what did?

Adults describing a late emergence often point to one of a small number of shifts, and they’re shifts in load rather than in wiring.

The demands went up. University, a first job with unstructured expectations, management, parenthood, a bereavement. The criteria’s own phrasing—social demands exceeding capacity—is a description of this.

The scaffolding came away. A parent who ran the logistics, a partner who handled phone calls, a job with a rigid routine, a school that supplied the structure. Remove the scaffolding and traits that were always there become load-bearing problems.

The strategies stopped paying. A qualitative study of 92 autistic adults built a three-stage model of camouflaging: motivations centred on fitting in and connecting with others; techniques combining masking and compensation; and consequences that included exhaustion, challenges to stereotypes, and threats to self-perception (Hull et al., 2017). Strategies that cost that much aren’t guaranteed to keep working forever.

Any of those can make a longstanding set of traits feel like a new condition. From inside it, “I’ve become autistic” is a reasonable description of the experience. It isn’t what happened.

Getting assessed when nobody noticed

If you’re heading into an assessment without a childhood paper trail, the useful preparation is developmental rather than current. School reports, a parent or older sibling who remembers what you were like, childhood photographs and home video, early interests, how you played, what you found unbearable. Where no informant is available, clinicians work from what you can reconstruct—the absence of an informant makes assessment harder, not impossible.

It helps to know what clinicians are listening for, which is rarely the dramatic material people prepare. Repetition and routine in how you played. What you did when a plan changed. Whether you had one or two intense interests and how you talked about them. Sensory things treated as ordinary at the time—the label in the collar, the hand dryer, the food that had to not touch other food. Whether friendships were made or joined. These are unremarkable details, and that’s the point: they’re the ones that survive in a family’s memory precisely because nobody thought they meant anything.

What an article can’t do is tell you whether your own history meets the criteria. Reading your childhood back through a hypothesis is difficult, memory is reconstructive, and that judgement is exactly what an assessment exists to make. For what that process involves, see what happens in an adult autism assessment.

One caution about the reconstruction. The risk of reading a hypothesis into ordinary childhood memories is real and runs in both directions—it’s as easy to dismiss a genuine early sign as normal as it is to recruit a normal memory as evidence. Bringing what you remember and letting someone else weigh it is more reliable than arriving with the case already built.

The question underneath “can autism develop later in life?” is usually whether a late recognition is legitimate. The diagnostic criteria answer that directly: they ask for early traits, and they anticipate that those traits may only become apparent much later.