emk wrote:Cainntear wrote:But that's not how the brain works, and we increasingly know it. The "critical period hypothesis" was controversial because it basically said that you couldn't master a language after a certain point in your life, and there were plenty of people who stood as counter-evidence, but people took it, evolved and built on it, and morphed itinto something more useful -- the critical period isn't about language acquisition in general, but about native language acquisition.
What is the current status of the "critical period" hypothesis? I last looked at it 10 years ago, and back then, my understanding was that:Have there been any new developments in critical period research in the last 10 years?
- There was extremely clear evidence of critical period for accents, which started closing around 6 and was closed for virtually everyone by 12.
- There were small but convincing differences between native and near-native performance with certain kinds of grammar rules. For example, even very strong non-native speakers of French with decades of practice tended to make a tiny number of gender agreement errors that didn't appear in native speech. This only seemed to apply to certain types of grammar rules, as far as I could tell.
- For most other aspects of language, there was no clear evidence of a critical period.
I last looked at it nearly as long ago, and at the time it was being presented kind of as a historical progression, going from an "everything" hypothesis to a more nuanced thing, and the sorts of things you mention are basically where critical period thought ended.
I seem to remember it being presented to us as general expectation of limits without specific intervention, and the thinking was that it had been incorrectly held as describing had limits with the notion that there was no point in trying to do better, cos you couldn't. I think it was being thought of as almost "control conditions" -- that research into teaching should be overcoming these limits, because otherwise you weren't proving your teaching technique was better than everything else.
Personally, I know a number of non-native speakers who moved to the US as post-docs in their mid-20s. In the case I know the best, they had maybe B2 conversational skills when they moved here, with frequent errors and odd word choice. Today, after decades of using English professionally, they have near-native English with vanishingly rare errors. And I know that they didn't do any English study during that time period—this improvement was organic. Near-native speakers often seem to be created by a process that looks a lot like childhood language acquisition. And it works best for voracious readers.
I hightlighted the post-docs bit, because they are clearly exceptional outliers whose brains are always active. Can we prove anything as generally true by looking at outliers? I would say something I heard in general school teaching: many teachers justifying themselves by looking at the success of the top of the class, but the top of the class are going to succeed whatever the teacher does and the bottom of the class are going to struggle whatever the teacher does; the only really valid measure is the average pupils -- the 33% that are neither top third or bottom third -- because they're the ones whose success depends on the teacher.
I imagine most post-docs are people who never truly needed a teacher, so the fact that they achieve great things without a teacher doesn't say much about the average learner.
Also—and this is clearly subjective—I do a bunch of things in colloquial French that I never learned "declaratively" and that I've never really tried to explain. I know how it should sound, and saying it differently sounds a bit weird. My judgement isn't nearly as good as a native speaker's, of course. (They have the advantage of a few hundred million words more input, plus most of them spent 13-20 years in school, and that actually counts for a lot!) But I do have procedural knowledge of French that I can't necessarily explain declaratively.
But that goes back to my point that there is always a mixture of things we learned declaratively and things we just picked up, because I was trying to say that the existence of things we were never told doesn't necessarily say that the act of telling us declarative rules is pointless.
Personally, I know I've had things that I never learned declaratively, and I guessed at them... but I very often guessed right.
To me, that implied that there was some kind of fundamental structure to languages that was tantalisingly just out of reach. Like, if it does X and Y with structure A, it will almost certainly do Z with A too.
This whole structure is next-to-impossible to describe, but it seems to me that what I'd built already implied the rest of it.
If I had make a guess, I'd say that second language learning from B2 through near-native levels is actually very similar to first language learning. It's not identical—even C1 students can benefit from explicit nudges about grammar—but I think the large majority of knowledge at those levels is often acquired as procedural knowledge
Language is said to be closely tied in to mirror neurons, so we would appear to understand language by a system of reconstructing the utterance and trying to experience what mental impulses would have caused us to say it.
Kids understand adults despite not having the same model as them, and they develop the model as they go, through understanding (Krashen's starting point).
Adults, however, understand each other despite not having the same model, but understanding does very little to change their model -- a Scotswoman who moves to Australia might start changing accent and phraseology around th edges, but will always talk in a noticeably Scottish way.
I.e. infants while speaking their native language are learning, but adults aren't.
To me, it seems likely that a "partially acquired" language model could fall into the trap of being effectively the speakers "own dialect", and then they can learn to understand "other dialects" without modifying their own.
I also think that it's possible to rely heavily on input even at lower levels, if someone is sufficiently determined to do so, and if they have some way to make input artificially comprehensible. But this may vary a lot between individuals. Some people have a good eye and ear for patterns. Even if they never try to explain those patterns in detail, they at least notice something interesting is happening. (The "noticing" hypothesis.)
Agreed. Possible, but not easy, and not optimal.