The Closing Door

Somewhere around your second birthday, a door in your mind began to close. For most of us, it stayed open just long enough. But what if, for some children, it starts closing early? Not slowly. Not partly. Early.

And what if the first sign is not what a child can do, but how fast they are learning to do it?

Here is a sentence. The dog bit the boy. Now the same words, swapped. The boy bit the dog. Same vocabulary. Same grammar. A completely different picture.

And the only place that picture exists is inside your head.

Neuroscientists call this prefrontal synthesis. Your prefrontal cortex reaches into the warehouse of memory, pulls out a dog, a boy and a bite, and bolts them into something you have never seen. It is how you understand the plate under the cup. How you follow a fairy tale about a three-headed horse. How you imagine anything at all.

It feels effortless. But some people never acquire it. Among autistic people, roughly thirty to forty percent cannot combine mental images on demand, even after years of therapy. They may learn thousands of words. But asked to put the plate under the cup, the picture never assembles.

The authors call it prefrontal synthesis paralysis. The question is why.

There are two stories that could explain it. Story one: a wall. Something, maybe sensory overload, maybe social withdrawal, stands in the way at every age. Learning still happens. Just slower.

Forever. Story two: a door. Autistic children start learning exactly as fast as anyone else. But their window closes early.

Here is the cruel part. Look at an adult, and both stories end in the same place: about sixty percent of the way there. You cannot tell them apart by where someone ended up. You have to watch the speed, while they are still a child. The paper draws it like this. Learning rate on the left, total ability on the right.

Wall on top, door below.

In twenty fifteen, a free language therapy app went online. Parents who downloaded it were asked to fill in a twenty question checklist, every three months, for up to three years. Does your child understand big apple versus small apple? On top of the box, versus behind the box? A cat ate a mouse, versus a mouse ate a cat?

Forty points in total. Each visit, the score moves. Fifteen thousand, one hundred and eighty-three autistic children. Ages two to twenty-two. Sixty-one thousand snapshots of a score going up.

But the researchers were not interested in the score. They wanted its speed. Points gained per year. The slope. The derivative.

They fit the simplest curve that describes almost everything that grows. Flat out at full speed. Then an exponential fade. Three numbers. R zero: how fast you start. T C: the age the fade begins.

Tau: how slowly it fades.

Think of a tap. R zero is the flow. T C is when someone starts turning the handle. Tau is how slowly they turn it.

Now let the data draw itself. The thin line is the average learning rate of fifteen thousand autistic children, at every tenth of a year. The thick line is the fit. It starts at five point nine points a year, at age two. And it is already falling.

Now the neurotypical children. At age two: six point one points a year. The same speed. Then the groups part ways. Neurotypical children hold full speed until about seven, when they simply run out of checklist. Autistic children begin fading at two point three years, with a tau of nearly five.

Split the autistic group by severity, and the door moves. Mild: the fade starts at three point two years. Moderate: at two. Severe: at one point four. Add up the area under each curve, and you get the destination.

Mild reaches one hundred percent, at seventeen instead of eight. Moderate and severe plateau below it, for good.

Same start. Earlier fade. Story two. The door. Not the wall.

Now the honest footnotes. These are parents' reports, not clinical tests, and parents can be optimists. The neurotypical group was just one hundred and thirty-eight children, recruited in parks in affluent neighbourhoods. The checklist is the authors' own. And a curve tells you when, not why.

On the why, there is a clue. In autistic brains, the genes that build synapses in the prefrontal cortex peak before age two. In typical brains, around age six.

A cortex that matures too fast may close its own window too soon. A hypothesis, not a finding.

But the stakes are not hypothetical. Autism is usually diagnosed around three and a half. Waiting lists push therapy to around four. By then, on these curves, the fade has been running for well over a year. For severe autism, more than two.

The same therapy at two and at four may not be the same therapy at all.

So zoom out. Every human mind is handed a world of parts, and a window in which to learn to combine them. How long the window stays open seems set, at least partly, by biology. What goes through it is set by whoever is talking to the child while it is open. The author has argued elsewhere that this same window is where language itself was born.

That is a far bigger, far more speculative story.

This paper's claim is smaller, and sharper. Somewhere right now, a two-year-old is learning what under means. Someone talked to you before your door closed. Pass it on.