A language learners’ forum

TopDog_IK wrote:If you aren't aware of German grammar rules while watching reality TV shows, what does it matter? You will pick them up soon enough through pattern recognition.

Will you though? One of my main arguments against Krashen has always been the observation that not everyone does pick up the grammar naturally -- many immigrant learners immersed in the language daily make mistakes for their entire lives.

Let's talk about how computer vision machine-learning algorithms detect cats. If you try to tell a computer a bunch of "grammar" type rules about cats -- "They have pointy ears, fluffy fur, four legs, a long tail", the computer will fail miserably at distinguishing cats from other animals or objects.

It is a favourite game of many machine learning researchers to see how they can confuse image recognition with impossible synthetic images.

The fact that this can be done shows that the image recognition has not actually learned what a cat is.

And of course image recognition systems can't be fed in reverse and expected to output an image based on a description. Spontaneous product from exposure to examples does not occur in artificial neutral nets.

When we want to train an image generating algorithm (eg DALL-E) it needs constant feedback on what it's done wrong -- far more feedback than a teacher could ever give a student in a lifetime. We have to do better than that.

But if you train neural networks on 100,000 labeled photos of cats vs other animals and objects, these NNs can classify cats at roughly a human level, much, much better than traditional algorithms utilizing grammar-style "rules". See what I mean?

But that's because the AI doesn't understand the rules - it doesn't possess general intelligence or understanding of language, which means the rules have to be impossibly explicit.

When you immerse with a second language, you are literally training the neural networks in your brain to recognize grammar and other language patterns. That's how it works.

Except that just like a computer can see a cat in an image of random eyes and fur, a human can understand a sentence without having learned all the grammar.

This is how we get to the position where someone can do practically everything in a language but still struggles with basic verb conjugation -- they can understand the sentence without ever having to notice the verb endings, so they don't acquire them.

TopDog_IK wrote:When you immerse with a second language, you are literally training the neural networks in your brain to recognize grammar and other language patterns. That's how it works.

Clearly that's not how it works when learning a foreign language, since no matter how much a person immerses (and I've done this on 4 languages in the country, I like immersion) quite a bit of grammar beyond very elementary grammar is not made clear at all.

To make some argument for not simply peeking into a grammar to straighten out minor problems along the way is just self-defeating rubbish.

Krashen's approach is a tool, not the entirety. No-one needs to turn this into a complicated pseudo-theory.

TopDog_IK wrote:Boy, when you show up as a new person to this forum, you better be ready for battle.

If you come at the king you best not miss!

A video posted here in a previous discussion with Olly Richards, the most interesting I have seen of Krashen.

https://forum.language-learners.org/vie ... 17&t=16113

I binged a few Krashen lectures this week and really cannot find much that I disagree with. He focuses very heavily on reading, as opposed to immersing with TV shows, but I really find myself in agreement with the vast majority of his ideas. The human brain is perfectly equipped to acquire grammar through immersion and pattern recognition.

TopDog_IK wrote: The human brain is perfectly equipped to acquire grammar through immersion and pattern recognition.

The human brain is also perfectly equipped to misunderstand the patterns too. And it's also perfectly equipped to learn the patterns before and get more out of it later on. You've addressed neither of these points. Nor any of the other shortcomings of Krashen's arguments that have been mentioned in this thread.

The word "perfectly" is not a good word to use here. It assumes a sort of fallibility of operations we know the human brain doesn't have.

Am I alone in thinking that 'Krashenism' typically affects people in their intermediate stages of their first language? At least this is what I've noticed in my interaction with his advocates.

einzelne wrote:Am I alone in thinking that 'Krashenism' typically affects people in their intermediate stages of their first language? At least this is what I've noticed in my interaction with his advocates.

Dunning-Kruger maybe?

The normal brain masters these declension endings perfectly:
- s
- es
- 's
- s'

The food for the "Krashen brain":