[You can also read this on my Substack, if preferred: Seeking Nisaba]
This one might get me labelled a heretic.
I started systemic advocacy from the point of view articulated in the oft-quoted paper from Snow and Juel:
Explicit teaching of alphabetic decoding skills is helpful for all children, harmful for none, and crucial for some.
I still think that’s true, but like many truths, it’s also nuanced. Or perhaps it is not really nuanced, as much as it is prone to being wielded without nuance, which often happens in Education.
Regardless, I’m feeling less positive about the rollout of structured literacy than I was initially. Not because it doesn’t work, but because of who it’s likely to not work for in the first instance: the ones for whom it’s most crucial. Let me explain by peeling the onion…
Not everyone is dyslexic
Mark Seidenberg has very publicly made the point that structured literacy was developed for dyslexics, and not everyone is dyslexic. There’s a discussion to be had about whether that’s an overreach (e.g. Direct Instruction literacy resources are a form of structured literacy, and they certainly weren’t ‘invented for dyslexics’) but I take his underlying point. He’s concerned about opportunity costs: what are the kids who take to literacy like a duck takes to water missing out on when they’re repeating work they don’t need to repeat? This isn’t a new position for Seidenberg (I touched on his take on that when I first wrote Merchants of Illiteracy in 2022), but he does appear to be becoming more vocal about it, and I would assume that’s because he’s seeing it becoming more of a risk, or an actual issue.
There’s also the concept of the ‘expertise reversal effect’ to consider: strategies such as granular instruction, gradual release, and lots of repetition can start to become ineffective or even detrimental once a learner has acquired a certain level of proficiency. Because, der. I mean, sure, it’s nice to have research that shows that this really is a thing, but the fact that we even need to point it out is rather odd.
But we do need to point it out, because we’ve all had or heard of experiences like that related by Seidenberg, when someone asked if we still need to teach phonemic awareness if a kid is reading fluently. (And, as an aside, just to undermine my own point with some more nuance, I’ll temper the exasperation likely evoked by that question by noting that there could be cases where PA work might be justified for a fluent reader, if their spelling is poor and suggestive of PA issues – some investigation would at least be warranted).
So, yes, things are often wielded without nuance in Education (not only in Education, but it is a pretty faddy sector), and wielding a highly structured low-variance curriculum without nuance could have unintended consequences. But this doesn’t mean that structured literacy can’t work. It just means structured literacy needs to be implemented well.
Which sounds like a cop-out. Because you could just as easily say “XYZ doesn’t mean that balanced literacy can’t work. It just means balanced literacy needs to be implemented well.”
And, to a degree, I’d say that’s actually true, because you can actually choose an effective adventure in a “choose your own adventure” approach. But it’s true to a far lesser degree than it’s true of structured literacy. Structured literacy is more likely to work for more kids in more classrooms than balanced literacy is, simply because balanced literacy incorporates theories of reading acquisition that we know have been proven wrong. Structured literacy is potentially prone to other problems (such as the expertise reversal effect and opportunity costs), but they are what I’d describe as implementation issues rather than issues with the underlying theory of instruction.
A well-implemented structured literacy approach can avoid expertise reversal effects and opportunity cost by ‘releasing’ a student once they’ve reached ‘escape velocity’ (Seidenberg’s analogy) and are effectively self-teaching. But, even then, they’d be best ‘released’ into other facets of structured literacy – fluency, sentence composition, narrative structures and so on – until they hit escape velocity in those domains as well.
No matter how well you implement a balanced literacy approach, it’s going to be based on factually incorrect assumptions of how literacy acquisition occurs, and by dint of its ‘balance’, is unlikely to feature the rigor required to quickly shift most novices into self-teachers.
This is, of course, Pam Snow’s ‘probabilistic thinking‘ around literacy instruction – what is more likely to get most kids over the bridge to literacy? It doesn’t mean structured literacy is perfect and, as I’ve written previously, we are nowhere near having an answer from science yet as to what is the absolute best possible
method of teaching a classroom full of kids how to read and write.
So, structured literacy is not perfect, but it’s going to do the best job out of the current choices. But there’s nuance within “the best job”, and I’m more worried about the kids at the bottom of the curve not getting enough than I am about the kids at the top facing opportunity costs.
Better and best is usually only better and best for some.
Despite being nowhere near having a scientific answer to knowing what the absolute best possible method would be, we can hypothesise about what the absolute best model of instruction would look like, based on background knowledge, observations, and ‘systems thinking’ – following processes to their likely conclusions, considering possible unintended consequences, and accounting for known constraints and how they’re likely to impact processes.
One of the things that has recently come onto my radar as background knowledge is the work of Matt Chingos and Grover Whitehurst, which found that shifting from a low quality curriculum resource to a high quality curriculum resource can have a bigger effect on student performance than a shift in teacher skill from the 50th to 75th percentile. Ollie Lovell and Ben Jensen kick this (and other great content) around in the ERRR #70 podcast (well worth a listen, thanks Bill Hansberry). Along with Ollie’s original position, I had always assumed that a good teacher would make far more difference than good curriculum materials. This work suggests this assumption is dead wrong (and worth comparing with John Hattie’s – arguably more dubious – effect size work that showed teacher expertise is way up there).
Indeed, I had even assumed that the driving motivation for some systems moving to evidence-based curriculum was to compensate for lower teacher expertise (especially in the case of Direct Instruction resources). I’m still not convinced that this assumption is wrong and, in a sense, it would be madness to not employ that compensation as a first step toward a system overhaul, given the superior effect. It almost doesn’t matter what the motivation was. Whether systems are attempting to ‘cut corners’ and improve resources rather than skills, or whether they’re doing that as an informed decision based on research like Chingos’ & Whitehurst’s, the end result should be the same: a better effect on student outcomes than would’ve been achieved by building skills.
So, what’s my problem? It’s with the gaps in this picture. “Better” and “best” outcomes in these discussions are about averages. And it’s a mathematical property of averages that it’s easy to improve them while further impoverishing those at either end of the curve. And even if those at the lower end of the curve are not actually further impoverished, an increasing average already leaves them further behind the average if they’re not enriched by the same degree.
In short, for example, if all your 20th-49th percentilers shift up to the 50th percentile, your 1st percentilers can contemporaneously remain at or below the 1st percentile (as opposed to compared to historical percentile norms) even if their skills improve.
Why would that happen?
Because it’s almost inevitable due to the way systems are rolling out structured literacy.
The major focus is on curriculum rather than teacher training (and should be, according to the research above, if viewed through a utilitarian lens). It’s also easier, not just more effective, to change materials than it is to create more highly skilled staff. But I’m yet to come across a ‘thought leader’ in this space who doesn’t contend that tier 3 intervention needs to be carried out by the best trained staff available. I’m also yet to meet an educational consultant who thinks Principals and higher-ups generally have a genuine understanding of how much work is involved in remediating an instruction-resistant reader.
I’m also yet to hear anyone explain satisfactorily how sticking to a program with exacting fidelity will remediate such a kid. I’m talking here about the kids with a psychometric profile that consists of, for example, extremely low rapid naming speeds, phonological awareness, orthographic processing, working memory, or all of these.
The Direct Instruction program, Spelling Mastery, for example, describes its “rate of instruction” as “twenty to twenty-five minutes per lesson, one lesson a day, five days a week.” There are 120 lessons in a Spelling Mastery level. In a school year of 200 days, it would be theoretically possible to move through a level and a half in a year, but in reality once ‘all the other things’ happen, that gets whittled back. It should also be noted that Spelling Mastery has 6 levels (A-F) and these are marketed to align with Grades 1-6, so that’s one level per school year. But even assuming that a school does religiously stick to 5 sessions a week and so traverses across levels within a school year, let’s look at what’s actually happening when this is first rolled out.
Kids do a placement test and are then streamed into their Spelling Mastery levels based on that. In a school that’s previously had a high variance Balanced Literacy curriculum, this will likely mean that you’ll have, for example, a few grade 5 kids doing Level C along with the grade 3 kids “at level” on Level C. Provided the teacher sticks to the script, those kids will get the same high quality instruction. But the elephant in the room (and he appears to be quite good at hiding, this particular elephant) is that those same kids are getting the same instruction with zero consideration for why some of the kids at level C are in year 5. That is, the placement test only tells you which words they can spell, and tells you nothing about why.
Here are a couple of possible explanations for the ‘why’:
- It could be that those grade 5 level C kids were, up until now, instructional casualties, and under DI they will make marked improvements. But they’re in year 5, and they now only have two years of primary school left in which to complete 4 years of content (Levels C, D, E, F) if they’re actually going to be remediated to grade level. Even under a best-case scenario of absolute fidelity, “one lesson a day, five days a week”, resulting in covering one-and-a-half Spelling Mastery levels of content in a school year, they won’t be beyond level E when they hit high school.
- It could instead be that those grade 5 level C kids actually had pretty good classroom instruction, but have significant lows in the subskills that contribute to literacy acquisition. They haven’t fallen behind because of the instruction, they’ve fallen behind because their particular profile leaves them highly resistant to good instruction. Let’s call them dyslexic. They’ll need more repetition, more granularity, more individualised teaching. But, instead, we’ve put them in what is effectively a great Tier 1 program at Level C along with the Grade 3 kids for whom that Level C class is Tier 1 or 2. In the Catholic Education Tasmania system, this kind of model is then labelled “Tier 3” (but only for the Grade 5 kid) and touted as best-practice, leading the State (which, scarily, it probably is).
Remember that this is the best-case scenario, where the program is implemented with exacting fidelity. And Spelling Mastery is pretty much the spelling program when it comes to evidence base. But even so, this implementation still leaves the kids in higher primary who have actually been identified via placement as being years behind, plodding along with the same content at the same speed as the kids who aren’t behind, years younger than them. How will they catch up?
They can’t – it’s a mathematical impossibility unless they have extra instructional time. For scenario 2 above, those kids are likely to also need extra attention and dynamic, diagnostic teaching if they’re to have any hope at all of catching up. Even if you were prepared to argue that deficits in RAN, PA, and orthographic processing are caused by poor prior instruction rather than being an inherent property of the kid, it will still require more effort to remediate those deficits through quality instruction. It will take more again if the kids are older.
So, I can’t conceive of how the same Level C script can have the same effect on a year 5 kid as it does on a year 3 kid. ‘Dosage’ required for consolidation varies markedly even within my own small caseload of diagnosed dyslexics.
For all these reasons, structured literacy rollouts will likely cater better to the middle of the curve than they do to the dyslexics for whom ‘structured literacy was invented’:
- Direct Instruction will work better for the middling readers and writers, since the script is likely to contain enough granularity and repetition for them, but not enough for the kids at the lowest percentiles on subskills.
- If a script/program/approach isn’t stuck to with fidelity (because staff don’t know enough about the pillars of learning to know why it needs to be adhered to, because we’re investing in materials, not knowledge), it likely won’t harm the ‘middlers’ so much as it does the lowest percentilers for whom that might be catastrophic.
- Many schools have now implemented a vast improvement in tier 1 instruction, and some are doing tier 2 well, but genuine tier 3 is still a mirage in the overwhelming majority. In Tasmania, many schools call tier 2 instruction tier 3, and I’ve only actually heard of one school that is apparently doing tier 3 well. In our own school, we were given something approaching tier 3 only after I threatened a discrimination complaint. So MTSS works comparatively well for the middlers at tier 1 and 2. The kids who need one-on-one targeted support mostly have a snowball’s chance, because the top of the MTSS pyramid isn’t really there yet.
- For the low middlers, it won’t matter so much if literacy ‘intervention’ is delivered by untrained TAs or music teachers, while for the lowest achievers this may well be an utter waste of everyone’s time. These kids need diagnostic, dynamic instruction.
- The expert literacy interventionists, meanwhile, are having responsibilities stacked on. In addition to coaching staff, they’re doing strategic work trying to operationalise the implementation of an actual MTSS which is properly staffed and progress monitored, and dealing with informed parents threatening discrimination actions. So they’re less available to do the actual intervention which the lowest percentile kids need – the highly attuned, diagnostic, task analytic, dynamic teaching that also requires time to reflect afterwards. Time they don’t have. And this is in the more functional systems and schools. In the less functional systems and schools there may not even be any leadership commitment to change, and teachers who are passionate about using evidence-based practice are doing all this entirely under their own steam, against resistance or apathy.
But the data…
And because this is all actually working pretty well for the middlers (of whom there are greater numbers), data will improve markedly despite the lowest percentile kids tanking because they’ve been ‘living off the scraps’. And that will be a juncture from which a number of outcomes are possible.
An individual’s improvement towards benchmark might be judged as evidence of response to intervention, which in isolation is then misinterpreted as meaning no further supports are necessary (an outlandish idea, I know, but I’m told it’s already happening).
Overall data improvements may be lauded as evidence of an effective system and, therefore, “enough”. System leaders and policy makers may settle for this and quit making changes while they’re ahead, treating the low percentilers as unavoidable and immovable outliers.
The lower end of the middlers improving, though, may lead to the goal that MTSS was always supposed to create: freeing up resources that allow a genuine tier 3 to be delivered. I suspect it would be a long and painful road to get there, though.
‘Living off the scraps’
Is that expression too harsh? There is quite the paradox here. In Tasmania, at least, these kids – for whom ‘structured literacy was created’ – are being sidelined by the very model that was designed for them, because it’s being pasted together in bits over a monolithic system that already has its own inertia. It is, so far, a bit of an in-progress, paper mache, structured literacy MTSS – patchy, sticky, soft, thin, and vulnerable – but it is bulking up and getting stronger, slowly. Whether it ever gets a solid centre or remains hollow will be ‘interesting’ to watch, and I do wonder if it would’ve been better to initiate a rollout more organically from individual schools which worked to plan and implement genuine and highly successful MTSS before replicating those in other schools.
But there it is, again: “I wonder if it would’ve been better…”. I’m talking about ‘better’ for some – the dyslexics – not for all. For the kids that just needed that little bit of explicit instruction to help them join the dots and crack the code, this rollout will be fantastic. And they’ll have fantastic skills, compared to the kids who still need much more than what the rollout is currently giving them.
Given the research says that upwards of 95% of kids can be taught to be literate, and cited rates of dyslexia vary in ranges like 10-20%, there’s undeniably a sense in which most ‘dyslexics’ can be said to be instructional casualties. What haunts me is that a possible result of these top-down, system-wide rollouts, is that we may also add “implementation casualties” to the lexicon – those who are further impoverished due to a system that currently, unintentionally or not, only caters for the middle or above. They’re getting high quality instruction, but they’re resistant to high quality instruction, so they need it at a genuine tier 3. It may seem counterintuitive that the kids for whom structured literacy is crucial, would not be served by a structured literacy rollout, but many unintended consequences do seem counterintuitive, until they happen, after which they suddenly appear blindingly obvious.
If that does happen (or, should I say, continues to happen), I’m not sure I’ll be able to console myself with the fact that our failure to improve the lot for dyslexics has been a result of succeeding to improve literacy for everyone else. Perhaps I need to polish my utilitarian lenses, but it does seem like a cruel irony. Should literacy reform be more in line with prioritarianism? I’ll leave that for ethicists to kick around, but perhaps prioritarianism is the last train stop at the end of the utilitarian line, because when we get to the end of the line there are less bums on seats, and that’s when the resources will be available to help.
That doesn’t make it an easier trip to endure, though.