Cavesa wrote:However, I disagree that accuracy comes only from speaking. Actually, speaking is an activity that goes on too fast for the accuracy (I often realize a mistake one second too late, as it has already left my mouth). All those tasks you mention require as much accuracy as you want them to require. If you do just the coursebookish listening activities with primitive questions, of course you don't need accuracy. If you challenge yourself to understanding completely, it is a different activity. Also, those inaccurate activities (like extensive reading and listening) can lead to accuracy, if you get exposed to the accurate content for long enough.
My view is a bit more subtle than I really said above.
There's several parts to it, but I suppose I'm coming at this from a teacher's perspective, rather than a self-teacher's.
First, with receptive skills (reading and listening), there is no way for a teacher to observe whether or not the student has genuinely processed the language using the right rules, rather than understanding just enough to get the meaning and then letting redundancy fill in the blanks. Any attempt to force the students to prove they've noticed a particular word draws far too much conscious attention to it, at which point you're not checking the language they've managed to automatise.
Then writing... well, I've seen lots of people who learn things in writing and then don't use it in speaking. The rules are consciously, explicitly known (declarative memory) but never internalised or automatised. Because writing can be done consciously, students can paper over the cracks without ever solving the underlying problem.
The reason speaking is so powerful is that a gap or mistake in the subconscious language model is readily evident. Speech must be done at speed, and even if you do self-correct, your initial error is apparent, so can be worked on to fix. (For me as a teacher, hearing both an error and an immediate self-correction gives me lots of information about my student's language.)
But even as a self-learner, I think speaking has an advantage here. I tend to self-correct when I make errors, because the time spent self-correcting is a nuisance and an inefficiency. My brain doesn't want to go through that process all the time, which leads to the brain trying to avoid the situation that triggers the self-correction. The same isn't true of self-correction in writing, because even in your own language, correction and revision of what you've written is part of the natural process (I couldn't tell you how many times I've deleted words and reformulated the sentences even in one paragraph of this message).
