My Year 1s can read the heart words but can't spell them
A Year 1 who reads said off the card and then writes sed in her recount is doing what the evidence says she should. Reading a word and spelling it run on the same knowledge, but spelling drags far more of it out of memory, so it lands later. Keep the sequence moving anyway: teach the parts of a heart word that already follow the code, dictate only what the class has had time to automate, and let the words cycle back through review instead of holding up the sounds.
Why spelling lands after reading
Linnea Ehri calls reading and spelling two sides of a coin. Closely related, highly correlated, but the amount of information a speller draws from memory is greater, because the speller has to produce multiple letters in the right order. Recognition is cheaper than recall. Meta-analyses put word reading and spelling at roughly r = .70 to .82: close, and not the same skill.
Her phase model does not put reading a phase ahead of spelling. Accurate word reading and correct spelling both arrive at the full alphabetic phase, once the major grapheme-phoneme relations are in place. Before that the child is partial on both sides, working from a few letter-sound links, often the first and last, so similar-looking words get confused and mail comes out as ML. Partial reading, partial spelling.
The writing system stacks it further. A sound has more possible spellings than a spelling has pronunciations. Measured at the rime of one-syllable Dutch words, consistency runs 84.5% going from spelling to sound and only 36.8% going the other way. English sits lower on both, 69.3% and 27.7%. Spelling also develops more slowly than reading even in orthographies considered transparent in both directions, so this is not your programme's fault.
Then there is the load. Spelling said means holding the sound sequence, pulling four letters back in order, and running the motor plan for each. Cowan put adult short-term capacity at about four chunks, and children well under it: around 2.3 in the first year of school, about 3.0 by Year 4. He reads that as a smaller store rather than a missing strategy, since the gap holds even when rehearsal and chunking are blocked.
What Prep, Year 1 and Year 2 actually ask for
The Australian Curriculum v9 answers the "should I expect both" question with a gradient, and it is gentler than most of us teach to.
| Year | What v9 asks for | Code |
|---|---|---|
| Foundation | read and write some high-frequency words and other familiar words | AC9EFLY14 |
| Year 1 | read and write an increasing number of high-frequency words | AC9E1LY14 |
| Year 2 | use knowledge of spelling patterns and morphemes to read and write words whose spelling is not completely predictable from their sounds, including high-frequency words | AC9E2LY11 |
| Year 3 | recognise and know how to write most high-frequency words | AC9E3LY12 |
Year 1's bar is growth, not a list to be mastered. Both tools for the awkward words are already in play by then: AC9E1LY13 asks for one- and two-syllable words with common letter patterns, AC9E1LY15 for grammatical morphemes. What Year 2 adds is pointing them at spellings that are not completely predictable from their sounds, the only place that phrase turns up anywhere from Foundation to Year 6.
Segmenting for spelling starts immediately, since blending and segmenting are reciprocal and the syllabus guidance weights the phonological component most heavily in the first two years of school. Patterns and morphemes as the deliberate attack on the unpredictable ones is where Year 2 lands.
The system won't flag the gap for you either. The Year 1 Phonics Check, which every state and territory has now agreed to run, is 40 words read aloud and nothing else. No spelling item. A class can look fine on it and still carry an encoding gap. (It also passes the child who reads accurately but slowly, and the papers repay reading by sound.)
Teach the regular parts, name the tricky bit
Most "irregular" words are mostly regular. Hanna and colleagues analysed 17,310 words: half spell correctly from sound-symbol correspondences alone, another 34% come out with a single error, and only 4% are genuinely irregular. Reading Universe puts around 85% of high-frequency words as fully decodable, rising towards 99% with comprehensive phonics. In would, only the oul needs remembering. In from, only the o. Take said: the s and the d follow the code, so the only part to learn by heart is ai saying /e/. Say the word, tap the three sounds, write the two regular letters, then point at the tricky middle and name it.
Little Learners Love Literacy defines a heart word as one children can't decode "at this stage", which is the right framing: un-decodable relative to the code you have taught so far, not permanently strange. Stages 1 to 4 carry about 15 of them, a smaller review pile than it feels like at 4pm.
Australian evidence points at writing the word as the active ingredient. Macquarie's Centre for Reading randomised 85 children in their first year of school across Look and Say, Look and Spell (the same routine plus writing the word), mispronunciation correction, and a control. Twelve irregular words, three ten-minute sessions on consecutive days, groups of five. Look and Spell and mispronunciation correction both beat Look and Say on reading the trained words, and the researchers put it down to actively processing the word's letters rather than looking at the whole shape.
Two caveats from the same study. Nothing transferred to untrained words, so each heart word still gets taught. And the errors that remained were a letter off the target rather than regularised guesses, which is progress you can mark as progress.
Dictation, word level before sentence level
Dictation is the bridge from decoding to encoding because the child retrieves the graphemes instead of copying them, and retrieval sticks far better than recognition. It is also high load, so pitch the support to how new the code is. A grapheme taught this morning can be dictated with plenty of help: say the word, segment it together, keep the grapheme card in view. Unsupported dictation is for code the class has had time to practise.
The shape that works: four or five words on whiteboards, segment with fingers first, write, then check grapheme by grapheme together. Add one sentence once the words are steady. The national Literacy Hub lesson flow ends the same way, sentence dictation using taught GPCs only, and Little Learners runs dictation from Stage 2.
Check whether it is the spelling or the pencil
Thirty-second test: ask the child to spell the word out loud, or build it with letter tiles. Right in tiles and wrong on the page points at the pencil more than the spelling. Treat it as a clue rather than proof, because tiles also put the right letters in front of the child, which is easier than recalling them.
Handwriting draws on the same working memory spelling needs, and until it is fluent it takes the resources higher-level writing wants. (We have covered why writing letters by hand beats tracing them separately.) In one study of five-year-olds' writing, handwriting fluency uniquely explained about 4.1% of the variance in written expression against spelling's 2.3%. Handwriting instruction carries its own effect size (0.66 against no instruction), though teaching legibility alone does not buy fluency.
Keep the code moving, cycle the words back
Waiting does not buy the spelling. Vadasy and Sanders randomly assigned five-year-olds to learn 10 GPCs or 15 over five weeks. The faster group learned the shared 10 just as well, and came out ahead on letter sounds, letter-sound writing and word reading (median d = 0.30).
Review does that work instead. Rosenshine's five to eight minutes at the start of a lesson, with a word introduced today coming back next lesson, again in a week, again in a month. Retrieval practice runs at g = 0.50 to 0.61 across the meta-analyses, and effortful retrieval beats easy retrieval, which is the case for dictating a word rather than showing it again. PhOrMeS sets a dosage target of at least 26 reads and 8 spellings per word, so even a programme that counts both expects the reading reps to run about triple.
Calibrate on how much comes back right. Rosenshine's benchmark for guided practice is roughly 80% correct, and Wilson and colleagues landed independently on about 85%. If your dictation set is coming back at half right, make the set easier rather than freezing the sequence. The extra dose for the kids who need it belongs in the groups you already run (one plan, different amounts of it).
Spelling is not a detour from reading either. Graham and Santangelo (53 studies, 6,037 students) found explicit spelling instruction improved spelling (0.54) and carried over to reading at about 0.44.
Choosing the dictation words each week
A dictation slot and a review pile you actually turn over will do most of this, plus some patience with a child who reads a word for a term before she can write it.
The weekly chore is choosing the words. Phonicspal pulls dictation words and sentences from the sounds your class has been taught, so the spelling practice sits where your sequence is up to. You can see how the teaching side works.
Curriculum expectations above are published by ACARA at australiancurriculum.edu.au. Year 1 Phonics Check detail from the Australian Government Literacy Hub.