Ever watch a three-year-old proudly sing the entire alphabet song, and then, thirty seconds later, stare blankly at the letter M on a cereal box?

Or watch a bright four-year-old write their name with half the letters completely backwards, leaving you quietly wondering if something is wrong?

Nothing is wrong. In fact, what your child is doing at that exact moment is pulling off one of the greatest cognitive feats in biological history.

Children are not little computers to be programmed or optimised. They are living, sensory wonders who learn through touch, warmth, play, and the sound of a loved one’s voice.

Yet when cognitive scientists recently tried to build an artificial “reading machine” (a large computer model trained on more than a million images), the experiment ended up teaching us very little about machines, and something profound about children: why early reading is so complex, why computers struggle with what a toddler does effortlessly, and what actually helps a young mind cross the bridge from shapes into stories.

1. The 5,000-year-old squatter

Here is a fact that changes how you look at a children’s book: the human brain was never designed to read.

Spoken language is ancient. Humans have been talking around campfires for hundreds of thousands of years; babies are born with brains instinctively wired to absorb speech from the surrounding air.

Writing, by contrast, is barely 5,000 years old. In evolutionary terms, that is the blink of an eye. There simply hasn’t been enough time for natural selection to evolve a dedicated “reading chip” in human biology.

So how does a child learn to read?

The brain performs an ingenious act of adaptation.

In a landmark framework known as neuronal recycling, the cognitive neuroscientist Stanislas Dehaene showed that learning to read gently repurposes a patch of the visual cortex originally evolved for recognising objects, identifying tools, and spotting shapes in nature.

When researchers scanned children every two months throughout their first year at school, they watched this unfold in real time. Within just a few months, a dedicated “reading headquarters”, the Visual Word Form Area, settled into place in almost the same spot in every child’s brain.

It didn’t erase their natural vision. It didn’t push out face recognition (in fact, their ability to recognise faces grew stronger over the same period). It simply loaned a small patch of visual hardware to help decipher lines on a page.

And why do letters look the way they do? Because when researchers analysed over a hundred writing systems across human history, they found something poetic: whether Chinese characters, Roman letters, or ancient scripts, the junctions where strokes meet (T, L, and X shapes) match the patterns found in natural scenes remarkably closely.

Our ancestors shaped the alphabet to reflect the natural world our eyes were already built to see.

Your child isn’t struggling to “learn how to see.” They are already a master at seeing. They are simply learning to aim that rich visual intuition at a new set of shapes.

2. Why b and d are not a mistake

If your four-year-old writes b instead of d, or reads p instead of q, you might worry about learning delays or dyslexia.

Take a deep breath: for a preschooler, this isn’t a defect. It is proof that their evolutionary vision is working perfectly.

For twenty-five million years of primate history, mirror invariance was a life-saving instinct. If a bird or a deer is facing left, and then turns to face right, it is still the exact same animal. A brain that treated a left-facing animal as a completely different creature wouldn’t make sense in nature.

Nature designed your child’s visual system to recognise that flipping an object horizontally doesn’t change its identity.

Reading is one of the very few activities in human culture that asks the brain to set that rule aside. Geometrically, b, d, p, and q are the exact same shape flipped like pancakes. To read them, a child must learn to gently inhibit that ancient mirror rule for letters alone.

And how does the brain solidify that subtle shift?

In a fascinating 2021 study of children in their first year at school, researchers gave the children short daily sessions practising mirror-letter discrimination. Some of them took an afternoon nap after practice; the others stayed awake.

Four months later, the children who had napped read about twice as fast as those who had not. Sleep and rest were quietly doing the delicate neural work that drills alone could never achieve.

3. The clumsy supercomputer vs. the wonder of the child

When scientists trained a modern visual neural network on 1.3 million photographs and taught it to recognise printed words, the computer eventually learned.

Out of its massive digital network, 33 artificial units stepped up to specialise in words. Switch off just a few of those units, and the machine became instantly “word-blind.”

While it was a triumph for computer science, the real takeaway was how clumsy the machine was compared to a young human child:

The AI supercomputer

Your four-year-old child

Needs 1.3 million images just to start

Needs a handful of loving exposures

Sees letters only as flat, cold pixels

Understands letters through hands, play, and movement

Cannot hear; has zero spoken language

Has years of rich speech to anchor each visual mark

Never rests or dreams

Consolidates discoveries during peaceful afternoon naps

A machine needs millions of data points because it is blind and deaf to the world. Your child has two unfair, beautiful human gifts: hands and voice.

The magic of the hand

When the neuroscientist Karin James scanned pre-literate preschoolers who had practised letters through freehand drawing, tracing, or tapping keys, she found that freehand drawing lit up the brain’s developing reading network more than tracing or typing did.

Why? Because when a four-year-old draws an R with a crayon, she draws it wobbly, imperfect, and different every single time. That very messiness isn’t an error. It is a rich, natural tapestry of shapes that teaches the brain what truly makes an R an R, far better than staring at a single rigid computer font.

The bridge of voice

A computer sees CAT and processes three arbitrary visual symbols.

A child sees C, A, T and connects them to spoken sounds she already uses every day to ask about the family pet. The letter on the page isn’t a cold piece of data; it is a living bridge to human conversation.

4. The alphabet song trap: why sounds come first

This brings us to one of the most common misunderstandings in early childhood: the alphabet song.

Singing the alphabet is a delightful memory achievement, but it teaches letter names (ay, bee, see, dee), not letter sounds.

If you ask a child to read the word CAT using letter names, they get:

“See, ay, tee” → “Seeaytee.”

That doesn’t sound anything like a cat!

To read, a child’s brain must connect the visual shape C directly to the sound /k/, so that /k/ /a/ /t/ effortlessly blends into CAT.

Letter names are helpful clues for letters like B (which begins with its sound, “bee”), but they can be confusing for letters like W (“double-u”) or H (“aitch”). If an app or toy only teaches names, it’s teaching children the labels on the tools without showing them what they can build.

5. What this means for everyday learning

When we built ABC Letter Meadow, we used this cognitive science as our guiding star, not to create a hyper-optimised drilling tool, but to build a calm, respectful sanctuary for early discovery.

Many children’s apps borrow mechanics from mobile casinos: flashing reward chests, ticking countdown timers, loud arcade sirens, and virtual gems. Those mechanics are engineered to hook screen attention, not nurture learning. They distract from the quiet magic of letter exploration.

We chose the opposite path:

  • Sounds first, never “buh”: Tap a letter and you hear its sound first, from a native-speaking voice actor who says B as a short, crisp /b/ rather than the muddy “buh” (the linguistic “schwa”) that makes blending words so confusing for young ears. The name button on the card plays the letter’s name, or you can switch on Letter names in the Parents area so the name follows the word whenever a letter is tapped.
  • Both capital and lowercase: An app that only shows capitals is only showing half a letter. Children learn the shape relationship by meeting both forms, so you can switch between capital and small letters.
  • Anchored to meaning: Letters don’t float in an abstract vacuum. Every letter is anchored to a real word and a gentle illustration, helping the visual brain and the spoken-language brain shake hands.
  • No pressure, no ads: Countdowns add pressure, and pressure gets in the way of exploring, so there are none. There are no virtual coins, no streaks, no tracking, and no subscriptions, and once downloaded it works fully offline: no internet is needed to play or to start the app.

What to do with the next ten minutes

None of this means a screen should ever take centre stage. Nothing in this research comes close to the effect of an adult, a physical picture book, and ten unhurried minutes together on the sofa.

An app should never pretend to be your child’s teacher. Its only modest role is to be a quiet, patient companion for the five or ten minutes they explore on their own, giving them the freedom to tap at their own pace, hear each letter’s name and sound as many times as they wish, and build joyful confidence.

So the next time your toddler pauses, furrows their brow, and mixes up a b with a d, smile. You are watching twenty-five million years of natural vision being gently invited to step aside for a bedtime story.

And that is nothing short of wonder.