Does teaching spelling work, and what does it transfer to?

Verdict

Spelling works, and it transfers backwards into reading rather than forwards into writing. That asymmetry is the whole topic and it comes from a single meta-analysis that had the discipline to report both halves.

Graham and Santangelo pooled 53 studies and 6,037 students from kindergarten to Grade 12. Formal spelling instruction against no spelling instruction or unrelated instruction: ES = 0.54. More formal instruction against less formal instruction: ES = 0.70 — meaning the choice is not only whether to teach spelling but how explicitly, and the explicit end wins, which is the same result the archive records for phonics and for explicit instruction generally.

Formal spelling instruction improved spelling, phonological awareness, and reading. It "did not translate into better performance on extended writing tasks."

Graham and Hebert arrive at the same transfer from the opposite direction. Their meta-analysis asks whether writing improves reading and finds that teaching students how to write improves reading comprehension, reading fluency and word reading. Two independent syntheses, run in opposite directions on overlapping literatures, converge on the same conclusion: code-level writing work pays off in reading.

moderate-support rather than strong-support because the corpus is a mixed-design meta-analysis (grade C), because effect sizes for spelling are measured on instruments close to the taught content, and because no independently evaluated randomised trial of spelling instruction anchors the estimate the way the EEF trials anchor the grammar and strategy topics.

What the evidence shows

The one genuine tension in the table is worth naming. Graham et al.'s elementary meta-analysis puts transcription skills — a category that bundles handwriting, spelling and typing — at 0.55 on writing quality, which looks like transfer to writing. Graham and Santangelo, looking at spelling alone with a purpose-built review, find no transfer to extended writing. Kim et al., restricted to K-3 with 166 effect sizes, find no significant transcription effect on any composition dimension. Two of three point to no transfer, and the two that do are the ones that isolated the question. The archive follows them.

The morphology results contradict two common assumptions at once. Morphological instruction is usually treated as an advanced topic for older or stronger readers. Bowers et al. find it is no less effective for younger students and that it brings particular benefit to less able readers — the opposite of both assumptions. They also find it works better combined with other literacy instruction than delivered in isolation, which is the same "embed it, don't unitise it" pattern the archive records for comprehension strategies and for grammar.

Hereditarian-lens assessment

Risk: low. The verdict rests on experimental and quasi-experimental comparisons in which children are assigned to spelling instruction or to a comparison condition, not on observational associations between how much spelling a family does and how well a child spells — which would be exactly the confounded design the archive's premise warns about.

The premise still does two pieces of useful work here.

First, it explains why the transfer asymmetry is the shape it is. Spelling and word recognition are substantially heritable with negligible shared-environment influence (Olson 2013). Spelling and word reading draw on the same underlying orthographic and phonological representations, so teaching one moves the other — that is near transfer, and near transfer is the class of effect the archive expects to be real. Composition quality is a further step out, loads on different capacities, and shows partly independent genetic influence from the speeded code-level measures in Olson's multivariate analysis. The failure of spelling to reach extended writing is therefore what the premise predicts, not an anomaly.

Second, it sets the expectation correctly. Explicit spelling instruction of the kind that works is floor-raising: it gets every child over a threshold that almost nobody crosses without teaching. It does not equalise spellers, and per METHODOLOGY it may well raise the measured heritability of spelling by removing an environmental bottleneck — which is evidence the teaching worked.

Nothing in this topic bears on g.

Boundaries & what critics say

• No independently evaluated randomised trial anchors the estimate. Unlike grammar and strategy instruction, where large EEF trials exist and shrink the meta-analytic numbers substantially, spelling has no such anchor. Given what happened to writing effect sizes in every other topic in this subject when an independent measure was applied, 0.54 should be treated as an upper bound.
• Spelling outcomes are measured close to the taught content. A spelling test after spelling instruction is near-transfer measurement by construction. The reading and phonological-awareness transfers are the more informative results precisely because they are further from the taught material.
• "Formal instruction" is doing a lot of work as a category. The meta-analysis contrasts formal with informal and more-formal with less-formal but does not specify a curriculum, a dose, or a sequence. Anyone quoting a required number of minutes is quoting a programme, not the evidence.
• The extended-writing null is a null, not a demonstrated zero. It is possible spelling instruction helps composition for children whose spelling is genuinely disabling and that the average washes it out — the same moderation hypothesis that reconciles the handwriting literature. Untested here too.
• Spelling time can be taken away, and one trial documents it. Two years of an intensive writing programme produced significantly less spelling progress. Whatever spelling instruction is worth, it has to actually happen.
• Morphology's evidence base is 22 studies with averaged effect sizes by outcome category rather than a single pooled estimate, so the magnitude is soft even where the direction and moderators are consistent.

Practical guidance

• Teach spelling explicitly and systematically. ES 0.54 against not teaching it, and 0.70 for the more explicit version against the less explicit one. This is cheap and it is one of the few places in the writing curriculum where "more structured" has a directly measured payoff.
• Justify it on spelling and reading, never on composition. Formal spelling instruction improves spelling, phonological awareness and reading, and independently the writing-to-reading meta-analysis finds teaching writing improves word reading and fluency. Those are the honest claims and two of them are measured on standardised instruments.
• Do not adopt a spelling programme as a writing intervention. The one transfer the purpose-built meta-analysis explicitly failed to find is transfer to extended writing.
• Add morphology early rather than saving it. Prefixes, suffixes, roots, and the spelling relationships between related words help younger students as much as older ones and help weaker readers most.
• Fold morphology into the literacy block rather than running it as a unit. It is more effective in combination with other literacy instruction than in isolation.
• Protect the time. A writing programme that runs for two years at the expense of spelling produced significantly less spelling progress. Spelling is not a rounding error in the timetable.

Open questions

• No independently measured, independently evaluated spelling trial exists. Every other topic in this subject that acquired one saw its effect size fall by a factor of three to five. There is no reason to think spelling is exempt, and no evidence either way.
• Does spelling instruction help composition for children whose spelling is the binding constraint? The average says no. Nobody has tested the moderation.
• How much, how often, in what sequence? The meta-analysis compares "formal" with "informal" and never reports a dose-response. This is the same gap as in every other topic here.
• Does the reading transfer persist? All estimates are end-of-treatment. The archive's general finding is that early literacy gains fade toward zero for typically developing children and persist mainly for the at-risk tail; whether the spelling-to-reading transfer follows that pattern is unmeasured.
• Where exactly does morphology belong in a sequence with phonics? Bowers et al. show it need not wait, but no study establishes the ordering.