rationalskepticism.org

katja z wrote:

Mr.Samsa wrote:
Yep, a lot of our creative and novel behaviors come about as a result of poor stimulus control, and the behavior can overgeneralise to other situations. Problem solving is an example of poor stimulus control, where we apply solutions from one situation to a completely novel problem. So I don't see why the same concept couldn't be applied to language.

Let me see if I've got this right. Problem solving comes about as a result of botched learning processes?? Now I'm thoroughly confused. Especially because I seem to remember you saying (on another thread) something to the effect that people can be trained to behave creatively. And because I thought humans were supposed to be excellent learners, but now you seem to be saying that we are great at not learning very effectively, which results in, well, effective behaviour in novel situations ...

Yes, pretty much. The process by which learning operates on is a selectionist process, like natural selection. Think of the "botched" learning mechanism as a form of fidelity, like a mutation rate in evolution, it's necessary for there to be errors in order for variation to arise - but obviously this error rate can't be too high, otherwise there would be no order or structure and the whole thing would collapse.

If we were "perfect" learners, and say we learnt that a red light means press the lever 50 times or whatever, then the associations we make would be to that specific stimulus and that stimulus only, and every time we saw it we would respond 50 times. When presented with an almost identical red light, but not the original, we wouldn't know what to do with it. We'd have to learn all over again that we have to respond 50 times in order to get our reward or whatever.

To bypass this cumbersome and expensive process, there are errors in learning. For example, our learning "fails" to perfectly capture every feature of the red light, so when we're presented with a similar stimulus, we respond as if it were the original stimulus. In addition, we also respond to stimuli which are similar but not identical to the original, so we might respond, to a lesser degree, to an orange or yellow light. This is an "error" or "failure" in learning, but it's necessary in order for us to function successfully. These errors form generalisations and allow us to make associations with similar concepts without having to explicitly learn them - like language for example, instead of learning a generalised grammatical rule and overapplying it to concepts (like "runned" and "drinked"), a perfect learner would have to learn the grammatical rule every time we learn a new word.

katja z wrote:

Could be! I don't know enough about it to know, but I like lateral thinking like that.

Yes, I like the results of poor stimulus control as well.

katja z wrote:

katja z wrote:But, with all that we can learn, I'd be surprised if we couldn't (re)train the ear to hear new distinctions (again), just like painters get very good with colours, or oenologists with taste, etc - I don't think we need to assume early exposure to other languages as the determining factor.

Probably true, and I think I had this discussion with some a while back.. Oh no, I think I was discussing teaching people perfect pitch. Same idea though, I imagine.

I've never thought of the two as connected, but it makes sense. Vowels, after all, are distinguished on the basis of signature frequencies (formants).

And, as far as I'm aware, notes have a kind of "critical period" like phonemes do too. I think Oliver Sacks discusses this idea in one of his books.

katja z wrote:But if you're right, if these late signers can use this specific area sometimes, why not always? Doesn't make sense to me.

It's just imperfect control, I imagine. If you suddenly take up a sport that required hand-eye coordination, you'd probably light up the same parts of your brain as a professional athlete would, but there would be less activation.

katja z wrote:On the other hand, we know that some (although rare) late learners do achieve native-like fluency in their second language (depending, however, on how you define "native-like fluency" - it isn't too clear a measure of language proficiency but what it certainly doesn't mean is "knowing the language inside out", because in this case many native speakers wouldn't fulfill the criteria!).

True, and I imagine those speakers would have similar activation rates as native speakers.