The science of spelling: how children actually learn to spell

Most adults learned to spell through some combination of weekly word lists, look-cover-write-check, and Friday tests. It mostly worked — for the children it worked for. For everyone else, spelling became something they were "bad at." Forty years on, we know enough about how the brain stores and retrieves written words to do better than that.

This is a tour of the research that actually matters for primary teaching: the cognitive science of how children learn to spell, what gets in the way, and what high-impact practice looks like in a classroom.

Spelling is not memorisation

The single most important finding from the last forty years of reading research: skilled readers and spellers do not store words as pictures. They store them as bound representations of sound, letters, and meaning — what Linnea Ehri calls orthographic mapping.

When a child reads the word night a few times with successful decoding, their brain binds the phonemes /n/ /aɪ/ /t/ to the graphemes n, igh, t, alongside the word's meaning. That binding is what lets them later spell night, fright, tight, flight — words they may never have explicitly studied. The "ight" pattern is filed under sound-and-meaning, not as a picture.

This matters because it explains a paradox every primary teacher has noticed: children can pass Friday's test on twelve memorised words, then misspell those same words in their writing two weeks later. Memorised word-shapes are fragile. Mapped words are durable. Teaching that builds mappings produces spellers who transfer to writing; teaching that builds shapes produces children who pass tests.

The four pillars of orthographic mapping

For a word to map properly, four things have to come together for the child:

1. Phonemic awareness. The child can segment the spoken word into its individual sounds. Without this, there's nothing for the letters to attach to. This is why phonemic awareness training has the strongest evidence base of any literacy intervention at Reception and Year 1.
2. Letter–sound knowledge. The child knows which graphemes (single letters or letter combinations) represent which phonemes. This is what systematic synthetic phonics teaches.
3. Successful decoding attempts. David Share's self-teaching hypothesis shows that each successful decoding of a word is a learning event. The child doesn't need to memorise — they need enough successful encounters with the word in context for the mapping to form. Estimates vary, but four to fourteen successful decodings is typical for a regular word.
4. Meaning. Mappings that include semantic information are stronger and more durable. A child who knows what journey means, has heard it in a story, and has decoded it successfully will spell it more reliably than a child who has only ever copied it from a list.

Why English is hard, and why morphology saves you

English is a deep orthography. In Finnish or Italian, letters map to sounds almost one-to-one — once a child learns the code, they can essentially spell anything they can say. English has roughly 44 phonemes encoded by around 250 graphemes, and the same letter combination can represent different sounds (through, tough, thought, though, thorough).

That depth has a cause: English is a layered language. Our everyday words come from Anglo-Saxon (house, water, bread). Our literary and academic words come from Latin via French (question, nation, society). Our scientific and abstract words come from Greek (photograph, democracy, chromosome). Each layer brings its own spelling conventions. Phonics alone can't explain them; etymology and morphology can.

This is why the DfE's statutory word lists shift in character around Year 3. Up to Year 2 the focus is phonics and common exception words. From Year 3 onwards, the lists are dense with Latin and Greek roots — and pupils who only have phonics as a strategy hit a wall. Teaching prefixes (pre-, sub-, trans-), suffixes (-tion, -ous, -ible), and roots (port, scribe, spect) explicitly is one of the highest-leverage interventions available at KS2. A child who knows that sub- means under and -marine relates to the sea will never misspell submarine again — and will be able to predict the spelling of dozens of words they've never seen.

How practice should be structured: the testing effect and the spacing effect

Two findings from cognitive psychology have changed how serious educators think about practice. Both apply directly to spelling.

The testing effect

Roediger and Karpicke's work (and a large literature since) shows that retrieving information from memory is a far more powerful learning event than re-reading or copying it. The act of pulling a word out of your head — even unsuccessfully — strengthens the memory trace more than any number of look-and-copy repetitions.

For spelling, this means low-stakes daily retrieval (a quick dictated word, a one-minute test) does more for long-term retention than thirty minutes of look-cover-write-check. Look-cover-write-check is recognition; dictation is retrieval.

The spacing effect

Cepeda and colleagues have shown that practice distributed across days produces dramatically better long-term retention than the same total amount of practice massed in one session. Three five-minute sessions across a week beats one fifteen-minute session at the end of the week — even though the total practice time is identical.

This is why the traditional weekly spelling cycle is so wasteful. Children typically receive a list on Monday, do little or nothing with it until Thursday night's homework, then take the test on Friday. The retrieval window is too narrow for spacing to do its work, and within two weeks most of the words are gone. A model that touches each word three or four times across a fortnight, with retrieval each time, produces durable spelling — even with no more total practice time.

Sentence context: why dictation beats word lists

When a child sits an actual spelling test in school — or writes a sentence in their book — they are not retrieving a word in isolation. They are retrieving it inside a sentence, with surrounding words shaping pronunciation, stress, and meaning. Practice should mirror that.

Practice that presents words in isolation builds a fragile representation tied to the isolated form. Practice that presents words inside a meaningful sentence builds a richer mapping, with stronger semantic and prosodic cues to retrieval. This is why dictation — the teacher reading a sentence aloud and the child writing it — is one of the most powerful spelling exercises available, and why it has been a staple of spelling instruction in countries with strong spelling outcomes for decades.

It's also why audio matters in any digital spelling tool. A word read in isolation by a synthesised voice is a poor stand-in for sentence-context dictation. A word read inside a meaningful sentence by a clear British-English voice is a near-perfect stand-in. SpellCast was built around this finding specifically: every word is read inside a curriculum-aligned example sentence in a natural British voice, so home and classroom practice mirrors the dictation children will actually face. Most spelling apps still read words in isolation — see the spelling app review for how providers differ.

What this means for teaching

Translated into classroom practice, the research points to a fairly clear set of principles. None of them are radical; together they're a meaningful upgrade on the typical Friday-test cycle.

• Daily, short, retrieval-based practice. Five to fifteen minutes a day of retrieval — not re-reading or copying — beats longer, less frequent sessions. Quick dictated words at the start of a lesson. A short app-based session at home.
• Distribute words across the week, not cluster them. Each word should be retrieved on at least three different days before being tested. A list given Monday and tested Friday, with no retrieval between, will mostly fail the spacing test.
• Practice in sentence context. Words in isolation are a poor proxy for the actual skill. Dictation — by teacher, by parent, or by a tool with sentence-context audio — should be a substantial part of practice.
• Teach morphology explicitly from Year 3. Prefixes, suffixes, and roots unlock far more words than rote memorisation can. The Year 3/4 and Year 5/6 statutory lists are essentially a morphology curriculum in disguise.
• Re-test cumulatively, not just terminally. Words from earlier in the term should reappear. Without re-exposure, even well-learned words decay.
• Make errors visible without making them shameful. Pupils need to see the gap between their attempt and the correct form to build the mapping. Low-stakes formats (self-marking, app-based correction, partner dictation) deliver feedback without the social cost of public correction.
• Connect spelling to reading and writing. Words that pupils encounter in their reading and use in their writing form stronger mappings than words seen only on a list.

How SpellCast puts the research into practice

SpellCast was designed by working backwards from the evidence above. Not every product decision is original — the research isn't. But the combination of choices is unusual in primary spelling tools, and each one is there for a specific reason:

• Sentence-context audio by default. Every word is read inside a curriculum-aligned example sentence by a natural British voice. Practice mirrors classroom dictation rather than the easier, isolated-word task most apps default to.
• Per-pupil adaptive spacing. Words a child is shaky on resurface across days, not all in one session. Words they've mastered drop out of frequent rotation but reappear later for cumulative review. The teacher doesn't have to engineer the spacing — it happens automatically per pupil.
• Retrieval, not copying. Every interaction is a dictated retrieval attempt. There's no look-cover-write-check mode because the research doesn't support it as a primary strategy.
• Full DfE statutory coverage with example sentences. Reception phonics through Year 6 statutory list. The Year 3/4 and Year 5/6 lists are organised so the morphological patterns surface — pupils encounter -tion, sub-, port as families, not as random words.
• Short, daily-friendly sessions. 10–15 minutes is the design target, because that's what the spacing literature supports. Long, infrequent sessions would be easier to build for and worse for outcomes.
• Custom word lists. Teachers can add their school's weekly list, science vocabulary, or topic words, and the same retrieval-and-spacing engine applies to those alongside the statutory list.

None of this is magic. It's an attempt to encode a few decades of cognitive-science research into the path of least resistance for a child practising at home and a teacher running a class. If you want to try it, schools get a free 6-week trial across the whole school — no credit card, full DPA pack returned within a working day. You can also see the retrieval approach in action with no login: the free two-minute quiz.

What the research warns against

• Pure rote memorisation. Producing a brittle, shape-based representation that fades within weeks. It can pass a Friday test; it doesn't transfer to writing.
• Look-cover-write-check as the only strategy. Useful for some learners as one tool among several, but it's recognition-heavy and lacks retrieval. Used alone, it under-performs dictation by a wide margin.
• Massed practice the night before. Most of what's learned won't survive to the test, let alone to next term's writing.
• Treating every misspelling as random. Most errors are systematic — phonological, morphological, or analogical. Teaching the underlying pattern fixes families of errors at once; correcting words one at a time doesn't.

Where this leaves us

The science of spelling is well-established and surprisingly actionable. Children learn to spell when they can hear the sounds in words, know how those sounds map to letters, encounter words in meaningful context, retrieve them across spaced sessions, and — from KS2 — understand the morphological building blocks the language is made of. None of that requires a million-pound research programme to put into practice. It mostly requires shifting the rhythm of practice from one big push at the end of the week to short, frequent retrieval across the week, with words in sentences and a deliberate eye on patterns.

Tools can help — the right ones automate spaced retrieval, sentence-context dictation, and per-pupil progression in ways a busy class teacher genuinely can't do by hand for thirty children at once. But the research doesn't say "use a tool." It says: build mappings, not memories; distribute, don't mass; retrieve, don't copy; teach the patterns, not just the words.

For UK primary teachers thinking about how this plays out in practice, see also: the best spelling apps for UK primary schools (2026), our free National Curriculum word lists with example sentences, and the Year 3 & 4 statutory list — where morphology starts to do most of the work.

Further reading

• Ehri, L. C. (2014). Orthographic mapping in the acquisition of sight word reading, spelling memory, and vocabulary learning. Scientific Studies of Reading, 18(1).
• Share, D. L. (1995). Phonological recoding and self-teaching: sine qua non of reading acquisition. Cognition, 55(2).
• Roediger, H. L., & Karpicke, J. D. (2006). The power of testing memory: basic research and implications for educational practice. Perspectives on Psychological Science, 1(3).
• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: a review and quantitative synthesis. Psychological Bulletin, 132(3).
• Bowers, P. N., Kirby, J. R., & Deacon, S. H. (2010). The effects of morphological instruction on literacy skills: a systematic review. Review of Educational Research, 80(2).