Maths might tell us how kids learn language

Misleading title

I was expecting some explanation of the process, rather than simply a description of the spurt.

Language isn't just vocabulary

The obvious flaw in this theory is that children do not just learn words - the acquire whole grammars. Even at the age of three they can substitute different vocabulary within a constant sentence structure, and have already got a grip on verb patterns, syntax, plurality and other structural features, not to mention precise punctuation and intonation.

Bilingual

And what does this reveal about bilingual children? What do these maths tell us about children learning two or more languages from birth? Is this language expansion on one plane or does it carry across to multiple planes? So many questions remain unanswered.

Non-story

I agree with Bob Appleyard, all the professor has done is describe the size of the vocabulary spurt, he's said nothing to explain how that vocabulary is acquired.

There's absolutely no story here at all.

And as the other poster said, grammar is arguably far more difficult to learn than vocab. A young child being able to tell that "the boy's tall" is a description and "the boy's book" is an object is immensely impressive, especially as "boy's" is pronounced (and spelt for that matter) exactly the same way in both cases. It's these kinds of abilities that distinguish a human being from the automated filing cabinets we know as computers.

Kids? Whaddyagonnado?

Don't all the offspring of close species demonstrate the same learning spurt? When we deal with language there's a bad hangover of the idea that it's language that truly separates us from the others of the animal kingdom. I'd like to see cross species studies to first get an idea of growth in learning rates during a similar age in other species then look at language learning rates as part of a broader context.

Pinker seems to be the torchbearer for Chomsky but there still doesn't seem to be a killer theory of language learning. I tend to go with a basic guideline of seeing our over sized, loosey goosey, associational cortex as demonstrating our ability to generate grammatical methods for symbol manipulation and communication. But, again, if I had my way more studies would look at vocalization in our close relatives then after we understand the basics of animal communication we can better get a handle on human language. For example if we hadn't taken to walking up right then it's likely we'd never have evolved to be able to articulate vocally on the wide range we do.

As noted by others the real fun part seems to be grammar as it reflects so much about our mental orientation. Grammar at some level could be a direct one to one transcription of our physical orientation.

3rd vote for non-story

Unless the article sells the theory sadly short, all that the guy has discovered is a mathematical way to describe the effects of total immersion on language learning at an age when the brain is still growing and - very importantly - neural connections are being extensively rewired.

Without major input from neurologists and psychologists, not to mention real linguists, the idea isn't going very far.

Not a non-story - but not the important story either

It is a curious result in many ways. It isn't that it is uninteresting, but as has been pointed out many times above, it makes little headway in understanding how we learn language. What it does do is establish a pair of _sufficient_ conditions that can explain the particular phenomenon - that of the shape of the learning spurt. What it does not do is establish the necessity of the conditions. So it provides some support to the idea that the learning of vocabulary occurs with the suggested underlying mechanism. But it doesn't prove it. It is quite possible that vocabulary comes about as a result different learning mechanisms that exhibit the same external behaviour.

Indeed this is the heart of the unsaid part of the story. It is is an example of a modelled phenomenon, where the emergent behaviour of a computational model seems to fit the known facts. Something that even has a real IT angle. That the model used is simple is good, that makes one more interested, and makes the idea presented more palatable.

So what do we learn? Well if we follow through, the results place some constraints on deeper models of language. They should be able to display an ability for concurrent learning of words. That isn't a trivial constraint. There are plenty of models of cognition that won't meet that. So the results provide some direction for other research. If further models conform to the constraints it is known a-priori that they can explain the observed learning spurt. What isn't shown is that if a new model does not conform to the constraints it won't explain the growth spurt.

The danger is that the model used may actually be as, or more, complex than the observed phenomenon, then it is simply a matter of searching for the right values of the free parameters to make the model fit the facts - and no science has been done at all. That is a serious problem with all such computational models.

The story is the continued growth in the use of emergent behaviour of computational models of complex systems to match the observed behaviour of real life systems. Emergent behaviour isn't new, it is 100 years since Einstein derived the gas laws from basic classical physics of elastic collisions between molecules, but matching up phenomena without a mathematical underpinning, and merely performing a computational simulation is new. Watching this impinge on different areas of science will be interesting.

Bilingual

Children in the language learning spurt stage are able to assimilate two or three languages equally as well as one. I live and work in Spain where a number of my aquaintances are members of mixed nationality marriages. Their children are frequently learning combinations of, for example:- Spanish, Catalan Spanish ( in Ibiza and Mallorca the schools only teach in the regional language), Russian and English. I am in my fifties, have been here five years and speak Spanish less well than a four year old, but I do always beat them in playground fights!

Misleading figures

I think the 50K vocab is incorrect. I have always believed that a vocab of 20,000 words is typical of a graduate (at least until the last 20 years of dumbing down). Offhand, I think English has 250,000 words, so suggestions that we use 50,000 seems incorrect.

I am lost for words...

Best wishes,

Martin

http://www.martinhwatson.co.uk/puzzle_news.html

Martin H. Watson

I am bilingual and never think about it

I speak English and Spanish and have since very young and I have never thought about why I can do it. I just thought it was my environment.

Vocabulary is just as important as grammar

Without vocabulary, you will not have much to say even if you know how to say it. I'd say vocabulary is more important than grammar. I was raised in the USA until age 11, when I followed my divorced mother back to France, learned French, and went through the French school system.

The irony is that I left the US school system in Junior High School and right when we had started studying grammar, and I was integrated in the French school system the year after they had finished studying French grammar.

So I know how to speak, read and write in two languages better than many, but I don't have the faintest idea why. I can tell someone if what he wrote is correct or not, but I can never justify my opinion. It's always a "gut instinct".

The one thing that helped me was my voracious reading appetite. I love reading all kinds of things, and that brought me a vocabulary much greater than the SMS-toting kids have today. It also probably showed me all the grammar stuff that I missed in school, but absorbed through the millions of words I have read.

So, in fine, I would say that vocabulary is more important than grammar.

Now, concerning this article, I am rather disappointed about the science of it. We are told that the researcher determined that his simulations demonstrated the mathematical tendency of learning better with large words. Great ! Now how were those simulations programmed ? Do we now have a precise mathematical model of brain function ? I doubt that, so what was his computer model programmed with and how is that relevant to the subject studied ?

It's easy to program a computer to give the results one wants to have. What guarantees that that has not been done here ?

Pascal.

Grammar

Pascal, I think grammar is important. Without grammar, vocabulary is just a heap of words. Difficult to say which is most important. But you learn grammar by practicing language (talking, listening, reading, writing), not by study. Just like you did.