This one is long, but it’s super worth reading.
(Interesting podcast episode on the topic of cognitive science of reading for background knowledge: https://brainsciencepodcast.com/bsp/2017/why-reading-science-matters-bs-136)
I’ll explain the punch line anyway: Primary schools in the USA are teaching children to read in a way that sets them up to fail. Some children will learn to read regardless, some won’t. The whole thing is an unfortunate interaction between settled science about how adults read, how they deal with new words they never read before, how children read, and how we should teach children. Adults can skip over words they don’t know and see if the text still make sense, barely look at a word beyond the first letter because they know which word to expect at that position in the sentence, and infer the meaning of unknown words from context. This is settled science.
The linked post is the story of different systems of teaching children how to read: Rote learning of whole words, phonics, context cues, or a mixed approach. The ideas about context cues vs phonics seem really obvious, and adults use a of context when reading text, they skim instead of sounding out, but the obvious conclusion (teach children to rely on context) is wrong.
If you teach children to read, you must not teach them to infer anything from context before you actually teach them to look at the letters, and that the letters mean sounds and the sounds form words, and the letters also form words. You need to teach that first, otherwise the children will struggle to read word, look for clues in the pictures or guess a word that fits into the sentence but they won’t associate the written letters with the sounds or the sequence of letters with the word as a whole.
The perceptual system in the brain is a Bayesian inference machine, and when you read enough, then your perceptual system will fill in the tables of word probabilities automatically.
But if you teach children whole words or phonics, but also teach them that they can sometimes skip over a word in first grade when the words are really simple, then it looks like they are learning to read but they actually don’t, and they will struggle to read even a simple newspaper article later in life.
Sure, if they are 12 and they don’t know the word “pericombobulations“, then you can tell them to skip the word or guess the meaning from context, and when you’re 51, and you read a newspaper article, you don’t need to craeflluy look at ervey ltteer wehn you can jsut mkae snese of teh graeetr wlhoe. But all this hinges on you being able to already know how to read, and it only works because you can look at the word “pericombobulations“ and say it out loud even if you never heard or read it before.
You can’t teach children the “algorithms“ or “heuristics“ adults use if you don’t make them develop a mental model of the territory. It’s like Lockhart’s Lament but for letters instead of numbers.
The Educational Upside Down is a parallel dimension where elementary school children are captivated by street signs and bored rigid by myths and tales of heroes. It is a dimension where early readers work out the relationships between the sounds of English and the letters that represent these sounds largely by being immersed in anodyne, specially written story books. Yet, weirdly, it is also a dimension where children have to be explicitly taught ‘comprehension strategies’ to understand what they read, such as activating their prior knowledge or deciding which sentence is the most important, and then must practice these strategies for the greater part of the school day. This is a dimension where knowledge of the world—that same prior knowledge that needs activating—is the last thing that it would occur to anyone to actually teach children in schools.
There’s a really interesting and important meta-level point to be made here about the nature of learning. I feel like I’m foggily groping around the edges of something big.
If nothing else –
– assuming, for the moment, that all the information in these articles is correct and that all the arguments are valid (and I do suspect that it’s all pretty on-target) –
– the mistakes being made by the advocates of cue-based literacy instruction are some really understandable mistakes.
They have a bunch of kids who can’t read well, and they’re not sure what they should be teaching those kids to do. They say “OK, let’s look at some really excellent readers, and see what those guys do, and teach the kids that.” Which is a very intuitively obvious “empirically-grounded” kind of plan.
Except that, as it turns out, what separates excellent readers from bad readers is a really solid foundation in basic-level technical skills, which aren’t visibly incorporated into any “strategy” because they’re so baked-in. If you want to produce more excellent readers you have to be mostly teaching them those technical skills. Worse yet: cargo-culting the actions of excellent readers, which is the thing you’re actually teaching, is the default strategy of confused incompetent readers and basically the Worst Possible Thing in terms of habit-formation.
…I have a sense that this precise pattern comes up a lot. I think people want to use “data” or “science” or whatever in order to find the Best Answer to their problem, and this leads to situations where they’ve identified Success but they don’t understand it, but they think that they’ve already done what they need to do in order to get a grip on the proceedings, and so their actual implementation plan amounts to cargo-culting.
I suppose that’s only to be expected, in a world where Success is almost certain to involve so many moving parts that you can’t possibly understand its mechanisms just by looking.
Aaargh – yes!
I'm facing this currently helping my daughter with basic arithmetic. She's learning how to add and subtract (whole) numbers with 2-3 digits in decimal numerals. I can see the problems my daughter's having with the whole thing. I can measure the 'inferential distance' between her understanding and knowledge of small cardinal numbers and the actually kind of sophisticated system we use to represent those numbers in our common "base-ten positional numeral system". And one key concept she's missing, and probably won't be introduced to for a while, is multiplication. '10' only really makes sense as 1 × 10 + 0 × 1. I'm hoping to get her an abacus as I think that might really help.
And the crazy thing is that multiplication isn't that hard. Her homework worksheets even visualize multiples of ten and a hundred using a visual representation of multiplication! In one case, the 'tens' were visualized as bundles of ten 'rods' – 'ones' were a single rod. A hundred was then visualized as a bundle of ten 'tens' bundles. Multiplication!
Later, on the same worksheet page, a 'hundred' was visualized as a square made of 10 × 10 little cubes.
And yet, my daughter, and her classmates, will probably learn enough to pass their class, and graduate to the next year. And they'll probably mostly be able to function mathematically – as much as anyone else is likely to demand of them.
And yet, almost no one could probably explain how the decimal numeral system works to represent whole numbers. And just one of the reasons why that's sad is that none of those people are likely to ever help anyone else really understand how it works either.
I'm going to go buy her an abacus ...
