Sight words vs heart words: what your kid actually needs to memorize

A five-year-old I know got stuck on the word "of" for about four minutes. He sounded it out the way he had been taught — /o/ /f/ — and produced "off." His mom, doing the thing every parent does, said, "Just remember it. It says 'uv'. It's one of those weird ones." He looked at her like she had broken the contract. Then he closed the book.

That moment is the entire reason the reading research community has spent the last decade moving away from the old approach to so-called sight words. "Just remember it" is not a strategy. It's a shrug. And we now have a much better answer.

Why English is like this in the first place

Before we talk about how to teach the tricky words, it helps to know why English has so many of them.

English is a layered language. The base is Old English, a Germanic tongue that arrived with the Anglo-Saxons in the 5th century. Then came waves of Old Norse from Viking settlement, then a heavy dose of Norman French after 1066, then centuries of Latin and Greek borrowings for science, law, and religion. Each layer brought its own spelling conventions, and English mostly kept them.

On top of that, between roughly 1400 and 1700, the Great Vowel Shift quietly rearranged how long vowels were pronounced — but the spelling system, by then locked in by the printing press, did not move with it. So "knight" still has a "k" and a "gh" because in Chaucer's day, those letters were pronounced. The word stayed. The sound left.

The result: about 50% of English words are fully decodable using basic phonics, another 36% are decodable if you know one or two extra patterns, and roughly 4% are genuinely irregular. That last sliver — plus the "almost regular" middle group — is the territory of heart words.

The old approach: memorize the picture of the word

For most of the 20th century, beginning readers were handed a list of "sight words" to memorize as whole visual units. Edward Dolch published his famous list of 220 words in 1936. Edward Fry expanded the idea in the 1950s with his "instant words." Both lists were based on word frequency — the 220 Dolch words make up roughly 50–75% of words in children's books.

The instruction was simple. Show the child the word. Say the word. Have the child say the word. Flashcard it until they recognize the shape. Repeat for all 220.

The theory was that these words were "irregular" or appeared so often that decoding each time was inefficient. The picture-the-word approach was supposed to build a fast visual lexicon.

It does not really work that way. And we now know why.

Orthographic mapping: the brain stores letters, not pictures

The neuroscience that changed everything came largely from Linnea Ehri, whose research on orthographic mapping showed that fluent readers do not store words as visual shapes at all. They store the letter sequence connected to the sound sequence and the meaning, bonded together in long-term memory.

That bonding process is called orthographic mapping. The brain takes a word, breaks it into phonemes (the sounds), matches each phoneme to its graphemes (the letters that spell it), and files the whole package away. Once mapped, the word is genuinely automatic — recognized in roughly 200 milliseconds, without sounding out, without guessing from context, without looking at the shape.

The implication is sharp. Even the words we used to call "sight words" get stored phonetically. The child's brain is doing phonics work whether we teach it that way or not. The question is whether we make that work explicit and fast, or leave the child to figure it out alone.

Reading Rockets has a clean overview of orthographic mapping, and the International Dyslexia Association covers the Ehri phase theory in its educator materials.

The new approach: heart words

Heart words come from structured literacy programs, most visibly the University of Florida Literacy Institute's UFLI Foundations curriculum, and from teachers building on Ehri's work. The core move is small but powerful.

Take a word like "said." A child in kindergarten who knows basic phonics can already decode the "s" and the "d." The trick is the middle — the "ai" makes a short /e/ sound, which breaks the usual pattern.

In the heart-word approach, you do this:

1. Write the word: s-a-i-d.
2. Tap each letter and say its expected sound: /s/ /a/ /i/ /d/. Notice the word is not "sayd."
3. Mark the irregular letters — the "a" and the "i" — with a small heart above each.
4. Tell the child: "These letters do something unusual here. You have to know them by heart."
5. Say the word together, then have the child say it while looking at each letter.

That's the whole technique. The child is not memorizing a picture. They are doing phonics on the regular letters and explicitly storing the one or two irregular bits. The brain gets the same mapping it was going to build anyway, only faster and with less guesswork.

The change in workload is real. Instead of memorizing 220 visual blobs, the child memorizes a few irregular letter-sound mappings per word, on top of phonics they already know. The cognitive bill drops by an order of magnitude.

Why the heart matters

The visual marker is doing real work. It tells the child two things at once: most of this word is normal, and exactly these letters are not. That precision is what was missing from "just remember it." Now the child knows what to remember and what to decode.

It also gives the child a felt sense of progress. The number of hearts on a word is the size of the ask. A word with one heart is almost regular. A word with two hearts requires a small leap. A word with three is rare. Children, including the five-year-old who closed his book, can handle a small leap when they can see it.

Spaced repetition does the rest

Knowing the trick is half the job. The other half is rehearsing it on the right schedule.

Heart words respond well to spaced repetition — the same retrieval-practice pattern that works for any memorized item. A typical schedule for a new heart word:

• 30 seconds after first teaching. Show the word again. The child reads it.
• Five minutes later. Show it again, mixed in with two or three other words.
• 24 hours later. Show it again the next day, at the start of the session.
• Three days later. Show it again. If still solid, the word is moved to a maintenance queue.
• Weekly maintenance. Show it briefly once a week for a few weeks until it is genuinely automatic.

Miss a step and the word slips back. Hit them on schedule and the word is mapped — not memorized as a picture, but bonded letter-by-sound-by-meaning in the way Ehri's research describes.

This is why heart-word work pairs naturally with structured phonics instruction. The phonics gives the child the regular letter-sound mappings. The heart-word routine handles the exceptions. The spaced-repetition schedule turns both into automaticity.

If you want to see how a similar pacing logic shapes the rest of a child's session, the science behind Lumi's pacing walks through how an adaptive tutor decides when to reintroduce a tricky word and when to leave it alone.

A walkthrough: teaching "said" the heart-word way

Old way: "This word is 'said.' S-A-I-D, said. Say it. Again. Again. Good. Tomorrow we'll do another one."

Heart-word way:

• Write said.
• "What's the first sound?" Child: /s/.
• "What's the last sound?" Child: /d/.
• "Good. The middle part — these two letters, a and i — usually say /ay/, like in 'rain.' But here, they say /eh/. That's a heart-word part. We mark it like this." Draw small hearts above the a and i.
• "Let's read it: /s/ /eh/ /d/ — said."
• Child reads it.
• Five minutes later, in the middle of a different task, the word reappears. The child reads it.
• The next day, it's the first word of the session.

By day three, "said" is filed. Not as a picture. As a sound sequence the child knows, with one heart-marked exception they have practiced enough times to retrieve in under a second.

Where Lumi fits

Lumi's reading sessions include a small task called HeartWordFlash. When the curriculum hits a word that does not follow the usual phonics pattern at the child's level, Lumi marks the irregular letters with hearts on screen, teaches the word out loud the way the routine above describes, and then schedules the word back into the queue at 30 seconds, five minutes, 24 hours, and three days. The child does not see the schedule. They just see the word again, at the right moment, until it is theirs.

That is the whole job. Phonics for the regular part. Hearts for the rest. Spaced repetition to lock it in. A five-year-old can do this. He cannot do "just remember it."

If you want a broader picture of the research base, the Science of Reading, explained without jargon is the place to start, and for parents whose children find decoding unusually hard, what early dyslexia looks like — and what helps covers what to watch for.

Try the Lumi beta.

— Tim

Image brief

• Hero image: An illustrated flashcard reading "said" in soft chalk lettering, with the "s" and "d" in cocoa brown and the "ai" pair marked with a small warm heart above each letter, on a cream paper background.
• Inline image 1: A simple two-column diagram comparing the old approach (a child staring at a flashcard of "of" with a thought bubble of question marks) and the new approach (the same word with a heart over the "o"). Placement: after the "The old approach" section.
• Inline image 2: A horizontal timeline showing the spaced-repetition schedule — 30 seconds, 5 minutes, 24 hours, 3 days, weekly — with small dot markers and a heart-word card riding along the line. Placement: in the "Spaced repetition does the rest" section.

Internal link suggestions

• Anchor: the Science of Reading, explained without jargon — links to /blog/science-of-reading-explained-without-jargon. Use near the end as the broader research primer.
• Anchor: the science behind Lumi's pacing — links to /blog/science-behind-pacing. Use in the spaced-repetition section as the pacing companion piece.
• Anchor: what early dyslexia looks like — and what helps — links to /blog/what-early-dyslexia-looks-like. Use in the closing paragraphs for parents whose child is struggling with decoding.

Editor's note

Two things for Tim to confirm. First, the HeartWordFlash task description — confirm the exact spaced-repetition intervals (30s / 5min / 24h / 3 days / weekly) match what's actually shipping in the beta; I used the schedule from the brief but the product may use slightly different windows. Second, the IDA outbound link points to the org homepage rather than a specific article — if Tim has a preferred Ehri-phase-theory page on dyslexiaida.org, swap it in. The opening "five-year-old I know" anecdote is composite/illustrative rather than a Remi-specific moment, so it stays generic; flag if Tim wants a real Remi scene swapped in instead.