Decodable or levelled readers: which book does which job
A levelled reader and a decodable reader are graded by two different instruments. A level describes the text on its own: sentence length, how common the words are, how often they repeat, how much the pictures carry, even how complex the letter-sound patterns are. Decodability describes the text against the code the reader has already learned, which in a classroom means the sequence your school actually teaches. So a book that is decodable for a class on one scope and sequence can be full of sounds another class has not met yet, and nothing on the cover will tell you which.
That second part is the bit most of us were never told directly, and it changes what you do with the box of readers in the corner.
What a level is actually computed from
Levelling systems rate features of the text. Fountas and Pinnell's own account of the gradient lists ten: genre and form, text structure, content, themes and ideas, language and literary features, sentence complexity, vocabulary, words, illustrations, and book and print features. The same page weighs sentence length, word length and the complexity of letter-sound patterns, so letter-sound information is genuinely in there. Say that plainly, because it is the strongest card the other side holds.
What is not in the ten, anywhere, is which of those patterns this reader has been taught. A level is a property of the book. It cannot be a property of the match, because the reader is not one of its inputs.
Quantitative formulas go narrower. Flesch-Kincaid runs on exactly two things, average sentence length and average syllables per word. Lexile uses sentence length and word frequency for texts above about 650L. Below that, across the early reader range, MetaMetrics' own documentation says the Text Analyzer runs nine variables and one of them measures decoding demand: easier texts "have words with fewer syllables and simpler sounds (e.g., 'net' and 'shop')".
So the early reader end of Lexile does look at decoding. It looks at how hard words are to sound out in general. Not at whether your class has been taught the graphemes inside them.
Levels are not a shared unit either. The NSW Department of Education warns that publisher numbers on decodable series are "only devised by the publisher and in no way are they linked or married to any other levelling systems".
Then there is how a level holds up under testing. Writing for Nomanis, Dr Jennifer Buckingham points at work from the University of Minnesota. In one study she cites, of 968 American second and third graders, measured against the district's own reading test, a one-minute oral reading fluency check classified students correctly 80% of the time. The Fountas and Pinnell Benchmark Assessment System managed 54%.
A second study from the same group had students read three different books, all rated at their instructional level by that system. The rating that came back, frustration or instructional or independent, matched across the three books only about 67 to 70% of the time, a kappa under .50 where .70 counts as strong agreement. Three books at one level, three different answers for the same child.
None of that makes a levelled book a bad book. It makes the number a rough sort rather than a fit test.
What decodability measures instead
Decodability is a relationship. The NSW Department of Education defines decodable texts as ones that "contain a very large percentage of words that incorporate the letter-sound relationships that students have been taught", and that "increase in complexity as the student learns more of the phonetic code". Taught is the load-bearing word in that sentence.
Which is why a series cannot simply be bought as a set. Five from Five puts its first criterion on the sequence itself: the order of grapheme-phoneme correspondences through the books "must be matched to the scope and sequence implemented in phonics instruction in the school". Two schools, two sequences, one box of books, different answers.
Dr Heidi Mesmer, whose research sits under most current definitions, treats it as a continuum rather than a label:
Instead of thinking of books as decodable or not decodable, we simply consider the idea of decodability to vary on a continuum, with some books having words that are more or less decodable to certain readers.
Her definition has two axes: how many of the words are phonically regular, and how closely the grapheme-phoneme correspondences in the text match the ones the reader has been taught. She sets no ideal percentage for that second axis, and researchers have disagreed on where the line falls. Collaborative Classroom's white paper puts a number on it anyway: once more than 10% of the words carry phonics or high-frequency words not yet taught, the text will most likely sit at that student's frustration level. Their judgement rather than a research finding, but a usable one, and note that an untaught heart word counts against you the same as an untaught grapheme.
A book marketed as decodable and stacked with sounds your class has not met is not decodable for them, which is a different problem from a book simply being too hard. The reverse holds too: a book can be fully decodable for a child and still too hard on the day, because knowing a sound and reading it quickly are not the same thing.
What to look for, book in hand
NSW's guidance is a list of things to look for rather than a stopwatch.
- Turn it over. Most publishers print the patterns a title teaches, and the one it targets, on the back. Match that against your own sequence before you read a page.
- At least 80% decodable to that reader. That is NSW's bar for a very high proportion of the text.
- The target grapheme gets a real run. It should turn up in a range of words and in different positions inside them, not once in the title and once on page six.
- The heart words are ones you have already introduced, so the book practises them instead of teaching them cold.
Alison Clarke's criteria land in the same place: one or two new sounds, plenty of chances to practise them, very little untaught phonics, and an order that follows a sequence.
Point three is where a lot of levelled stock falls over for a beginner, and not because the books are poor. Clarke pulled apart one widely used early levelled title and counted 31 different sound-letter correspondences in it, most appearing once or twice. Fine for a child who already reads. Nowhere near enough repetition for a child still consolidating a pattern.
Then hand the book to the child and listen to a page. Around 95 to 97% accuracy is the band for teaching something new, and 98% and up is where fluency practice belongs, which we have covered separately.
The levelled readers already in your book room
Nobody is binning a thousand dollars of stock, and nobody needs to.
A caveat first, because it cuts both ways. A University of Newcastle systematic review located 1,865 published articles on decodable and levelled readers and found only 19 met its criteria for inclusion, concluding the evidence base for both falls short of what is needed to make informed decisions. A 2024 meta-analysis of 16 studies by Murphy Odo found decodable texts had a small to moderate measured effect on word reading and pseudoword decoding, and worked best as one part of a broader programme rather than the only resource.
There is a real trade-off in the research too. Mesmer found first-year students reading decodable text matched to their phonics lessons were more accurate and needed less help from the examiner, and read the less decodable levelled texts with greater fluency. Dense decodability can slow a beginner down, because they attend to every grapheme rather than reading in phrases.
So the useful question is which instructional slot a book fills.
- A child still building the code needs the text for teaching and practising a new sound to be decodable for them. Rereading familiar decodable texts is then the fluency work, and in a rotation that rereading is one of the routines a table can run without you.
- Once a child has the code for what is inside a book, levelled stock does plenty of jobs: rereading, vocabulary, comprehension talk, home reading, a story you read together.
Even Timothy Shanahan, who has argued at length against matching children to levels (his American Educator piece is titled "Limiting Children to Books They Can Already Read"), names the beginner as the exception. Students who cannot yet read at around a high Year 1 level should be working in texts with high decodability and controlled vocabulary.
The NSW Department of Education's own video guidance frames decodables the same way, as "another tool in our toolkit that we would use for a very specific learning purpose for a short period of time". A slot, not a replacement philosophy. In a rotation, that slot sits inside one plan with different amounts of it.
Talking to a colleague who wants everything matched to a level
Keep it on what is being measured, not on which scheme is better. That second argument goes nowhere, and it is not the actual disagreement.
Two questions usually do it.
What decision is this number making? For grouping a class roughly, or tracking growth across a year, a level is a serviceable sort. For choosing the book a child will practise Tuesday's new grapheme in, it cannot answer, because the taught code is not one of its inputs.
What does this child do with the book in front of us? Hand it over, listen to a page, count the errors. An accuracy check settles it in the room, and it is hard to argue with a child reading aloud.
If the curriculum comes up, it is on your side. Version 9 of the Australian Curriculum removed its references to predictable texts and to the three-cueing system, and asks Foundation students to read decodable and authentic texts. That is the document talking, not a position on anyone's books. We have written up what V9 changed inside a phonics block if it helps to have it in one place.
The harder version of this conversation is the one where decodables are handed out full of sounds the class has not been taught yet. Same measurement error running the other way, and the back cover catches it.
Checking a whole box of books
Running that check across twenty books is an afternoon, not a job for the five minutes before a group. Phonicspal does that bookkeeping for its own texts. It holds your class's taught sounds and heart words, and every word in a lesson, story or printable is checked against that list first, so what comes out is decodable for the class in front of you rather than for an average one. You can see how that works or start free.
A level describes the book. Only the reader in front of you settles whether it fits.
Decodable-text definitions and the 80% guidance above are published by the NSW Department of Education.