Nonspeaking Autism: What Is Actually Happening in Your Child's Brain.
You already know your child understands you
You know it from the way they move toward the door when you mention the beach. From the look that crosses their face when you say something you probably should not have said in front of them. From a hundred small moments that never make it into a report.
And yet the words do not come. Or they come rarely, or unpredictably, or nothing like the way everyone keeps promising they eventually will.
Somewhere along the way you have probably been told that your child's language centres are affected. I want to talk about that, because the imaging has moved, and what it now suggests is both more precise and, I think, considerably more hopeful.
This is a longer read than most of what we publish. You have almost certainly done more reading on this than most people who advise you, so I am not going to simplify it.
Let Us Start With What The Words Mean
Nonspeaking and minimally speaking are not the same as speech delay, and neither one tells you anything about your child's intelligence.
Speech delay implies the same road, travelled slower. Minimally speaking usually describes a child who by school age has fewer than around thirty functional words or phrases. Nonspeaking describes little to no spoken output at all.
There is also a good chance childhood apraxia of speech is in the mix, which is a motor planning difficulty distinct from autism. Your child can have both. Working out whether they do changes what we would do about it.
Here is the part I most want you to hold onto. Speech output is a poor measure of comprehension, and a poor measure of thinking. Almost every standardised cognitive test requires a verbal answer. Which means the children whose abilities are hardest to see are the ones we most reliably underestimate. If you have ever sat in a meeting where the numbers on the page did not match the child you know, that is not you being an optimistic parent. That is a measurement problem.
What Speech Actually Requires
Speaking looks like one act. It is not. It runs across a network, and each part does a different job.
Wernicke's area, in the left temporal lobe, turns sound into meaning. If your child does not seem to register a spoken request, this can be why, and it has nothing to do with their hearing.
Broca's area, in the left frontal lobe, builds the motor plan for speaking. This is the gap between knowing what you want to say and having a set of instructions for your mouth to follow.
The arcuate fasciculus is the white matter tract running between the two. Meaning in at one end, speech out at the other, and this is the cable carrying the message across. It is one of the last tracts to finish myelinating, and imaging studies in autism keep finding it thinner than expected. When this is the constraint, your child understands and wants to answer, and the message simply does not arrive.
The lateral cerebellum is the part almost nobody mentions to you. It connects to the opposite side of the brain, so the right cerebellum supports the left sided language regions. Its job is timing, sequencing and prediction. When it is involved, speech tends to be effortful and unrhythmic rather than absent.
The insula helps coordinate the physical mechanics of getting sounds out in the right order.
The Finding That Changes The Picture
Earlier this year, researchers published the first resting state MRI study of nonspeaking and minimally speaking autistic children in Brain Communications. They were looking for what was wrong in the language regions.
They did not find a damaged region. Broca's area was there. Wernicke's area was there. Structurally, the parts were intact.
What they found was underconnectivity, spread across the whole language network and reaching into the insula and hippocampus, and it tracked with the children's clinical scores. DOI
I want to be careful not to oversell what this means, so let me say it plainly.
Your child's speech network is not broken. It is undercoupled. The pieces are present, and they are not talking to each other well enough.
That distinction is not a comforting reframe. It is a clinically important one, and it cuts both ways.
It means treating one spot is unlikely to be enough. If the problem is spread across a network, whatever we do has to be as well.
It also means the thing that is limiting your child is a dynamic property rather than a fixed one. Damaged tissue stays damaged. Connectivity does not work that way. It changes in response to repeated, targeted input, and it changes most readily while a brain is still developing.
Why We Assess Before We Treat
From the outside, four very different problems look almost identical.
A child who does not answer a spoken request might not be decoding the language. Or they might be decoding it and not attending to it. Or attending and unable to assemble the motor plan. Or able to do all three, with the message failing somewhere in transit. Same behaviour in the room. Four different neurophysiological stories, and four different responses from me.
So I measure rather than guess…
A QEEG and ERP assessment gives me direct readings rather than inferences. Coherence and phase lag analysis shows how your child's speech regions are actually communicating, and in which direction. It is the practical clinic equivalent of the connectivity research above. Alongside it, I can look at how your child's brain responds to sound automatically, without them having to look, point, press a button or say anything.
That last part matters more than it sounds. Every assessment your child has faced has probably required them to produce something in order to be counted. This one does not.
The purpose is not to hand you a thick report. It is to answer one question about your child specifically. Where in the chain is the constraint.
How Light Fits Into This
Photobiomodulation uses light on the scalp to change how the tissue underneath produces energy.
The most studied target is cytochrome c oxidase, the last enzyme in the mitochondrial energy chain. It absorbs light in the red range and again in the near-infrared range. The light appears to lift inhibitory nitric oxide off the enzyme, letting electron flow resume and ATP production increase, with flow-on effects to local blood supply and inflammation.
Why that would matter for language follows straight from the connectivity finding. Networks that are struggling to stay coupled are metabolically expensive to run. Give the tissue more energy to work with and you give the network more capacity to hold its connections.
Not All Light Is The Same
This is the part that gets flattened in most explanations, and it matters, because wavelength determines what the light reaches and what it does when it gets there. Different wavelengths hit different biological targets and produce genuinely different responses. They are not interchangeable, and more is not better.
| Wavelength (band) |
Primary biological targets and mechanisms | Key biological responses | Research favoured clinical applications | Depth of penetration | Evidence maturity |
|---|---|---|---|---|---|
| 808 nm Infrared Deep |
|
|
|
Deep (several cm) | ● ● ● ● ● High Strong clinical and preclinical evidence |
| 975 nm Infrared Superficial to mid depth |
|
|
|
Superficial to mid depth (1 to 2 cm) | ● ● ● ● ○ Moderate to High Growing clinical evidence |
| 638 nm Red Surface to mid depth |
|
|
|
Surface to mid depth (up to about 1 cm) | ● ● ● ● ● High Well established clinical evidence |
| 528 nm Green Surface |
|
|
|
Surface (less than 1 cm) | ● ● ● ○ ○ Moderate Emerging clinical evidence |
| 405 nm Violet Surface |
|
|
|
Surface (less than 1 cm) | ● ● ● ○ ○ Moderate Emerging clinical and preclinical evidence |
For the speech network specifically, red and infrared are doing the work. Red is absorbed strongly near the surface. Infrared travels further and more of it reaches cortex. That is exactly why we use red first and infrared second at each site, so we get the benefit of both absorption peaks rather than choosing between them.
Green and violet are a different conversation, and I want to be straight with you about where that conversation sits. Both are surface wavelengths. In autism specifically there is interest in violet for its proposed effects on Purkinje cell density and GABA signalling in the cerebellum, and in green for oxygen release from red blood cells. Those ideas are biologically plausible and worth watching. They are also, at this point, preclinical and early. Anyone presenting them to you as established is getting ahead of the data, and I am not going to.
What I Can And Cannot Tell You About The Evidence
I would rather you hear this from me than discover it later.
The most relevant study is a 2025 open-label trial that targeted Broca and Wernicke areas directly. Over ten weeks, the researchers reported a seven-point improvement on autism rating scales, alongside decreased delta and increased gamma and beta power on EEG, with the size of the EEG change tracking the number of sessions. DOI
A randomised sham-controlled trial in younger children found a difference of just over seven points between the treatment and sham groups, with no moderate or severe adverse effects. DOI
Now the caveats. These are small studies. Most are open-label. Some had device manufacturer involvement. And not one of them used expressive language as its main outcome, which means I am reasoning across from general autism scores toward a language effect. That is a real gap, and I am not going to paper over it.
What keeps my interest is the EEG. Rating scales in unblinded studies of children are vulnerable to everyone involved hoping very hard. Brainwave power is much harder to hope into existence, and the fact that the change scaled with the number of sessions is exactly the pattern you would expect if something real were happening. That is encouraging. It is not proof, and I will not present it as proof.
What We Actually Do
We treat the network, not a single spot.
Where. Lateral cerebellum, Wernicke's area, along the arcuate fasciculus, and Broca's area.
Light. Red first at each site, then near infrared for depth.
Sides. Left hemisphere language regions, paired with the right cerebellum, because the cerebellum serves the opposite side.
Time. Usually two to four minutes per site, longer over Broca's area when getting words out is the main sticking point, and adjusted as we see how your child responds.
Order. Broad coverage of the whole network first, then narrowing in once the assessment and your child's early response show us where to concentrate.
At A Glance
| Region | What it does | What it looks like when constrained |
|---|---|---|
| Lateral cerebellum | Timing and sequencing | Effortful, unrhythmic speech |
| Wernicke's area | Sound into meaning | Requests do not seem to land |
| Arcuate fasciculus | Carries the message across | Understands, wants to answer, nothing comes |
| Broca's area | Motor plan for speaking | Few words despite clear comprehension |
| Insula | Mechanics of articulation | Difficulty coordinating the physical act |
One Last Thing
Light is one part of what we do. It sits alongside QEEG and ERP mapping, tDCS, tACS, tRNS, tVNS, PEMF and the Rezzimax. What we use depends entirely on what your child's assessment shows, not on what we happen to own.
And neuromodulation supports the brain. It does not teach language. The children who do best have good speech pathology and occupational therapy running alongside, with communication supports already in place.
Please hear that last part properly. Giving your child a way to communicate now, in whatever form works, is not giving up on speech. Everything we know points the other way. A child who can make themselves understood builds language, gets less frustrated and finally gets to show people what they know, and all of that supports the very network we are working to strengthen.
If you want to know which part of your child's speech network is the real constraint, rather than guessing from the outside, that is where we would start.
Bliss Jackman is the Principal Practitioner of The Togetherness Project and provides neurotherapy and family therapy services in Hawthorn, Vic and Fremantle, WA