Slashdot Mirror
How Infants Crack the Speech Code
scupper writes "Infants learn language with remarkable speed, but how they do it remains a mystery. New data shows that infants use computational strategies to detect patterns in language, according to UW's Dr. Patricia K. Kuhl in the Nature article "Early Language Acquisition: Cracking the Speech Code" [PMID: 15496861]
Interesting excerpt from the article: 'There is evidence that infants analyse the statistical distributions of sounds that they hear in ambient language, and use this information to form phonemic categories. They also learn phonotactic rules -- language-specific rules that govern the sequences of phonemes that can be used to compose words.'"
Yes, in a manner of speaking. They first learn what the language is supposed to sound like. The abstract tells us how the infants form words and sentences, but it doesn't tell us how they map the sounds to their meanings/contexts. Maybe the main article goes into more detail. I think the word/sound->meaning/context mapping would be interesting to study.
There are computer programs that can recognize words (voice recognition), but how many programs can (with a large rate of success) recognize the words and map them to their meanings, or context? The point about the neural net is also interesting. It would seem that the brain is programmed to understand a certain language better. Does that mean that people who have learnt a certain language, can learn a similar language easily? The article seems to suggest that if the neural net is built in a certain way, it might be easy to learn similar sounding languages, but a language with a very similar grammar, but different sounds might be difficult? Would be interesting to pursue and find out...
Vivin Suresh Paliath
http://vivin.net
I like
There are various theories. In the generative tradition, humans are born with a vast amount of knowledge about language. In that sense they already know "grammar" before they learn individual words. On the other hand, to work out the settings for the various innate parameters they have to be able to segment into words, so many linguists would probably say that "grammar" acquisition runs alongside lexical acquisition. For more information, read anything by Chomsky.
Other theoretical traditions would say that there is no innate grammar, but rather that learning a language consists of learning statistical patterns which are represented through neural activation patterns. For them, grammar will follow lexical acquisition. Other argue that the lexicon is effectively the grammar. For more information, read anything by Elman or Bates. Both the latter have articles online which can easily be found by googling, but I'm a lazyarse and can't be bothered to do it.
The critical period theory, that a child can only acquire a first language until the beginning of puberty, has been confirmed in many case studies. For obvious ethical reasons, these experiments cannot be set up intentionally, but in cases such as a severely abused child who was never exposed to language until about age 10, a woman who was deaf until a surgery when she was 30, the peopl e who have not yet reached puberty are still able to learn a language normally, and the rest are not. I strongly recommend reading The Language Instinct by Steven Pinker if you are interested. Pinker also discusses differences between learning a language and learning other things. For example, in most other things children learn, they see exactly what is done and then mimic it. However, learning language also gives a child the ability to create a sentence he has never heard before. Additionally, language is learned with no formal instruction, whereas other skills must be taught actively.
Nobody is saying that adults can't ever reach fluency. The claim is that as you get older your ability to learn languages decreases rapidly. If both you and a five year old are immersed in a foreign language environment, she will (barring a huge exception) inevitably end up speaking the language better than you. You need to distinguish between fluency and *native* fluency. Adults who are able to achieve fluency that is comparable to that of a native speaker are very rare, and while the limits vary from person to person there will almost always be a wall past which one cannot progress.
For example, children start having trouble being able to hear the difference between sounds that are non-constrastive in their native languages as early as 18 months. If you poke around in the literature on developmental psychology, you'll probably come across stories about "Jeanie", a strange case of a child who was basically locked in a dark room for her entire childhood. Despite sincere attempts, there was no success in teaching her anything that resembled a human language.
There was also a recent study done on Nicaraguan Sign Language(a form of sign language that's being invented as we speak by deaf children who had no previous access to sign language). It's an interesting case, because the language originates from a school in Managua so every year a fresh group of first year kids are newly exposed to it by their older peers. Over the years NSL has evolved substantially from a more iconic gesture-like system to one that is begining to demonstrate hallmarks of universal linguistic properties, such as the building of hierarchical phrase structures and the serialization of complex ideas into separate words. This has happened rapidly, so the younger kids sign quite differently than the older ones. The older kids, and especially the young adults who were among the first NSL speaking classes, have retained the more primitive gestural components of the language and are basically stuck in that pattern, more or less unable to augment their signing skills with the newer features. The conclusion reached by the study is that not only do young children have a better time learning language, but they also seem to have a brain that's specially adapted to the creation of language from scratch, an adaptation which does not appear to be similarly shared in mature adults. Cool stuff.