Over the past couple years, a consensus has emerged among business, philanthropic, political, and many education leaders about how we should be teaching young children to read. Inspired by the hugely popular podcast “Sold a Story,” by the journalist Emily Hanford, most U.S. states passed “science of reading” laws, mandating that literacy be taught through systemic, deliberate, phonics-based methods. “Sold a Story” debunked a set of popular ideas about how kids learn to read that had been ubiquitous in schools for years, most notably associated with Columbia University Teachers College’s Lucy Calkins’s Units of Study curriculum. Calkins promoted “balanced literacy” or “whole language” approaches, that, critics say, didn’t do enough to teach kids the fundamentals of phonics, and instead emphasized memorizing sight words and using context clues to guess unfamiliar words.
The “science of reading” on the other hand, relies on decades of, as the name suggests, scientific research about reading, language, and learning. In particular, the science of reading approach calls for explicit, systemic, phonics-based instruction to teach kids how to read. As it’s implemented in schools, it’s often associated with an approach called “Structured Literacy,” which was developed by the International Dyslexia Association. Because dyslexic children by definition have great difficulty with decoding and word recognition skills, the structured literacy approach was designed to be very deliberate and thorough, breaking the task of reading words into bite-sized pieces, and introducing these pieces to students in a systemic, step-by-step process. With all the attention placed on the science of reading and phonics-based instruction, structured literacy, which was developed to remediate reading gaps for dyslexic children, ended up being used for general education purposes.
In other words, when young kids are being taught how to read today, they are all being taught to read as if they are dyslexic.
As you might imagine, some folks are now saying this might be a problem – that we might have, with the best of intentions, gone a bit overboard in the rush to align with the science of reading. And one of the people sounding the alarm is Mark Seidenberg, a psycholinguist whose book Learning at the Speed of Sight, helped launch the science of reading revolution in the first place.
In a talk Seidenberg gave last year, he goes through how structured literacy was developed and came to be widely adopted, and talks about why he now sees this as a problem. The talk is rather long, but it’s well worth the time.
The core argument of his talk is that to teach reading well, we need to get the balance right between explicit and implicit instruction: the things we teach explicitly, and the things we should leave for kids to pick up on their own. At the risk of oversimplifying things, the idea is that in the days of balanced literacy and whole language instruction, the pendulum fell too far in the direction of implicit instruction. Teachers and curriculum developers viewed reading as something that children picked up naturally, with enough exposure. This, of course, is not true. As Seidenberg notes, there are some rules, some basic skills, that need to be explicitly taught – the sounds letters typically make, how various letters typically sound together, etc. Without this explicit instruction, kids will have a hard time learning how to read.
However, when states began adopting science of reading legislation, and therefore the structured literacy approach, teachers began to be taught, and therefore taught kids, all the rules. Phonemes, diphthongs, schwas, fricatives. If you don’t know what these words mean, you’re in good company. As Seidenberg notes, most people don’t know what these words mean, because they were taught to read without them. But in today’s classrooms, students are being taught these words, and all of these rules and theories that come with these words, that used to only be reserved for language scholars. Seidenberg says the result is that today young children are met with a “barrage of explicit instruction.”
All of this is a problem, according to Seidenberg, because students don’t only learn from explicit instruction. Rather, they learn all sorts of things implicitly – on their own, without being formally taught, during various other, more enriching and active learning activities. He notes that the widespread adoption of structured literacy operates on the assumption that what’s good for dyslexic kids is good for all kids, or at least won’t be harmful to typically developing kids. But this is not true – because teachers have a limited amount of time with students, and there are huge opportunity costs to teaching all these individual components of reading.
In other words, the cost of doing all this explicit instruction on reading rules is all the other skills and knowledge students are not learning during that time, not only in reading but across disciplines. This is a problem because to be a great reader you not only need to know how to sound out words, but need to know what all these words mean, how they relate to other words, what is inferred by these words, what might happen next. These are skills largely gained not from explicit instruction, but from lots of actual reading. As Seidenberg says, “Over-teaching the components of reading eats into time for other learning activities, such as reading itself.” He notes that of course some explicit instruction is needed to get early readers off the ground, but that “…most of the knowledge that supports reading isn’t actually learned from instruction. If you try and teach it all, you are interfering with or taking away from other learning opportunities that kids need.” Indeed, this may be one reason why we have seen some early grade gains in reading, where improvements may be more easily achieved through explicit instruction, but gains stall out in later grades, where higher-order reading skills are assessed.
This is really important because, Seidenberg says, what allows learning to really take off occurs on the implicit side, or what he calls “statistical learning,” or learning from patterns. Only after repeated exposure to lots and lots of language do you begin to recognize patterns, setting you up to learn more things. This is the kind of learning that happens without you really knowing about it. He notes that this is the kind of human learning that large language models, artificial intelligence, was based on. It’s a core feature, a sort of magical feature, of how humans learn. But if we only barrage kids with explicit instruction, we may be preventing kids from really taking off, and letting their implicit learning systems run. “The goal of reading instruction isn’t to get the child to a fifth grade reading level,” Siedenberg says. “The goal of reading instruction is to get the child to escape velocity.”
The close of Seidenberg’s talk is worth quoting in full:
“The science of reading is committed to explicit instruction, as a way to ensure that everyone succeeds,” but “it isn’t realistic, it creates unnecessary hoops, and it takes time away from other learning opportunities, including various kinds of reading itself. Is the approach effective? We don’t know. Is it efficient? Definitely not. And…the opportunity costs are huge.”
In other words, of course phonics are incredibly important. You can’t teach someone to read without teaching them some basic phonics rules. And under previous approaches, like balanced literacy and whole language, we were doing an insufficient job of teaching phonics. But we have way overcorrected. And the negative impacts this could have on our young children’s education – in terms of opportunity costs, and limiting student potential – are substantial.
