Speech delay or autism? What actually tells them apart

Our daughter had a speech delay. It is not what led us to the diagnosis

We want to start here, because it cuts against how this question is usually framed.

Before we knew anything about autism, our daughter did have a speech delay. We asked about it. A doctor in the family told us what we also found repeated everywhere online: that speech delay is not uncommon in children under three, and that many children catch up.

That was reasonable advice. It was also true. And so the speech delay was simply not the thing that made us seek an assessment — other things did. Her speech therapy came afterwards, once we had the diagnosis.

What a speech or language delay actually is

Worth separating two things that get muddled. A delay means language is developing along the usual path but slowly — “not in line with the stages expected for their age”. A disorder means it is not following the typical pattern at all, and persists.

Late language emergence — the clinical term for a late talker with no other diagnosed difficulty — is common, and the reassuring part is real: roughly 50% to 70% of these children are reported to catch up to their peers by late preschool or school age.

Note what that also means. A meaningful minority do not, and “most children catch up” is a statement about a group, not a prediction about your child.

NHS services also make the point that a language delay “may be the main difficulty or part of another difficulty such as hearing, global developmental delay or autism” — a delay is a symptom, and it has more than one possible cause. Hearing is worth ruling out early.

What clinicians actually weigh

These are not a test you can run at home, and we would ask you not to treat them as one. They are the kinds of things an assessment considers, offered so the process feels less opaque.

The central logic is this. Some communication behaviours do not require speech at all — showing, directing someone’s attention, coordinating gaze, using facial expression and gesture. Because they do not depend on language level, they separate the two questions cleanly. A child with a language disorder and no autism “would be expected to show many such compensatory communication strategies”. In autism these are, by definition, more limited.

So the question is less “how many words?” and more:

• Does the child route around the missing words? Pointing, leading you by the hand, fetching a thing and putting it in your hands, miming.
• Do they share things purely to share them? This is a real and subtle distinction. Pointing to request something is often intact in autism; pointing simply to say look at this with me is the one more specifically affected.
• Joint attention — following your gaze or your point, and drawing your attention to things in return. It usually emerges around nine months and is well established by twelve.
• Pretend play and imitation. Limited pretend play and limited imitation are much more common in autism than in language disorder alone.
• Response to name, and use of eye contact and facial expression.

Be careful with any of these in isolation. The NHS is explicit that “every autistic person is different and many autistic people will not show all the signs of autism, or signs may not always be obvious”. These are patterns clinicians weigh together, not switches.

They are not either/or

This is the part the “versus” framing gets most wrong, including in the title of this article.

Language difficulties are common in autism, not exceptional. Among children with neurodevelopmental conditions such as autism or ADHD, the prevalence of language disorder is estimated at 30% to 50%. Studies of eight-year-olds on the autism spectrum have found language disorder in anywhere from 23.6% to 63.4%, depending on the study.

And the picture varies enormously. Around a quarter of autistic children have typical or even exceptional language ability by age five, while others use limited verbal communication into adulthood.

So “is it a speech delay or autism?” often has the answer: both. Finding one does not rule out the other, which is why a speech-therapy referral is not a substitute for a developmental assessment if other things are also concerning you.

Rough milestones — and why they are rough

Parents ask for numbers, so here they are, with the caveat the sources themselves attach: language development follows a broadly fixed course, “although variation does occur in terms of the timing of specific stages”.

• First words — most children at around 10 to 12 months.
• Two-word combinations — typically between 20 and 24 months.
• The commonly used flag for late language emergence is fewer than 50 words and no two-word combinations by 24 months.

Even that flag comes hedged: individual differences in these skills before age four “are not highly predictive of later language outcomes”. Milestones are a prompt to ask a professional, not a verdict.

Losing words is different from being slow to gain them

If your child had words and then lost them, mention it explicitly. Clinicians treat that differently from a child who has simply been slow to start.

Language regression — losing acquired language in early childhood — is described in the research as a distinctive feature of autism. How common it is depends heavily on how you measure it: around 22% in population-based samples, about 34% in clinical samples, and around 41% when based on parent surveys. We give you all three rather than pick the scariest or the mildest.

One genuinely reassuring finding: regression “does not necessarily foreshadow worse long-term developmental outcomes”.

You do not have to know which it is to start helping

If there is one thing on this page we wish we had known earlier, it is this.

You do not need to resolve the question before getting help. Speech and language support helps a child communicate whether the eventual answer is a delay, autism, both, or neither. As one professional body puts it: “You don’t have to ‘wait and see’ before seeking support.”

Singapore’s own early-intervention system is built on that principle. As a Singapore Medical Journal review of the local ecosystem puts it, because a young child’s developmental trajectory “is still fluid, it may not be possible or desirable to make a confirmatory diagnosis of a specific disability condition” — support is arranged around functional need, not a label. In practice that means EIPIC can begin on a paediatrician’s assessment that a child is at risk of developmental delay.

In our case therapy came after the diagnosis. It did not have to.

What to do with all this

Nothing here can tell you which it is, and it is not meant to. The honest position is the one the National Autistic Society states plainly: “the only reliable way to find out whether someone is autistic is through an autism assessment.” Even experienced clinicians treat this as a careful differential, not a snap judgement.

What it can do is sharpen what you bring to a doctor. Instead of “she isn’t talking yet”, you can say what she does instead — whether she points to show you things, whether she brings things over to share, whether she follows where you look. That is far more useful to a clinician than a word count.

Then take the ordinary next step: our guide to what to do if you suspect autism covers the free developmental screening and how to raise it, and the signs of autism in toddlers checklist goes age by age.