I just listened to this podcast a few weeks ago. What really struck me was just how obviously wrong the Guided Reading curriculum and three cueing theory was even just on its face, and how many people went along with replacing phonics with Guided Reading despite that.
For example, the podcast recounts a Guided Reading lesson where the teacher covers up the words the kids are supposed to be reading with a piece of tape and tells them to guess what the word is without seeing it. That feels so obviously dumb that there's no way anyone could have been fooled into thinking that was a superior method of instruction than phonics, right? Surely the podcast has to be exaggerating? If not, that's a level of stupidity that's actually high enough to make me angry.
Same thing with the idea that people read based on context (three cuing). Like, sure context can be a component of reading, but it's trivially disproven as the primary mechanism with a simple string of random words like: nanny overlying identify crinkly eats reunion. Is that hard to read? Obviously not, so clearly context isn't that important for strong readers, and teaching kids to guess words based on context rather than teaching them to actually read the words is dumb. You shouldn't need a scientific study to figure that out.
> covers up the words the kids are supposed to be reading with a piece of tape and tells them to guess what the word is without seeing it.
That produces kids that don't read, in the normal sense of the word; rather, they synthesize a text based on guesswork, and confidently declare that it says what it doesn't say. My daughter is a primary teacher, and has complained about this. It happens.
It seems completely barmy, to me.
Incidentally, it sounds a lot like what ChatGPT does.
It's true, masked-language-modelling has been hugely successful in natural language processing. So a priori I wouldn't have guessed it would work so poorly for humans and that phonics-based teaching is so much better. But the evidence is clear.
Just because MLM can work doesn't mean it does work in all contexts. We get to change the software (and hardware although that has been less relevant IMHO) behind ML models as much as we want, for humans we need to adapt the training methods.
Your example sentence was incredibly hard to read for me. To be fair English isn't my primary language. But still, I think it could be that different people read in different ways, so some depend more on context than others, etc.
I think they've extended the meaning of "reading" to include a certain level of "understanding what you just read". Your nonsense sentence can't pass the latter test because there is no understanding to be gleaned, but it's not at all obvious that that should how we view reading exactly for that reason.
Correct, but I don't know if they've extended anything. When you say someone reads at the 12th grade level, it has never meant that they can superficially sound out every word, it means they can understand what the text is trying to say.
For example, the podcast recounts a Guided Reading lesson where the teacher covers up the words the kids are supposed to be reading with a piece of tape and tells them to guess what the word is without seeing it. That feels so obviously dumb that there's no way anyone could have been fooled into thinking that was a superior method of instruction than phonics, right? Surely the podcast has to be exaggerating? If not, that's a level of stupidity that's actually high enough to make me angry.
Same thing with the idea that people read based on context (three cuing). Like, sure context can be a component of reading, but it's trivially disproven as the primary mechanism with a simple string of random words like: nanny overlying identify crinkly eats reunion. Is that hard to read? Obviously not, so clearly context isn't that important for strong readers, and teaching kids to guess words based on context rather than teaching them to actually read the words is dumb. You shouldn't need a scientific study to figure that out.