01 September 2001
Language Acquisition: Even more miraculous than you might think
The more we learn about how infants learn language, the more miraculous it seems. Babies begin to figure out how the sound system of their language-to-be works, in fact, well before they've even learned to speak or understand a single word. Believe it or not, there's even good evidence that language acquisition begins before birth. When a baby is born, she's already more attracted to her "mother tongue" — literally, the language spoken by her mother — than to the sounds of other languages. Already in the womb, the fetus has listened to the prosody, or intonation, of its mother's speech, and comes into the world primed to focus on that language as the one it's most important to learn.
After an infant has learned to identify the prosody of her mother tongue, she has to learn the specific sounds. Many people have heard the claim that infants can produce the sounds of any language in the world, and only later lose that ability once they've figured out which particular sounds are important in the language the infant learns. Actually, this claim isn't quite true. All infants find some speech sounds easier to produce than others. Vowels are easy, as are consonant sounds such as "b," "n," or "d." Fricatives ("f," "th," etc.), affricates ("ch"), and liquids ("r," "l,") are universally harder to produce, and are therefore relatively rarely uttered by very young children.
On the other hand, a very young infant does have a remarkable ability to hear all the sounds, and differences in sound, used by any of the world's 3,000 languages — despite the fact that most of those different sounds are ones their own mother tongue doesn't use. What happens then is that the infant learns to pay attention to the sound differences that are important in just the language spoken around her, and to ignore ones that are not. These changes in an infant's ability to hear the differences between speech sounds actually involve changes in the wiring of the infant's brain, with the result that it becomes harder and harder to hear the sound distinctions that only other languages consider important. That's one reason why it's so much harder to learn to speak a foreign language fluently, and without an accent, at age 15 than at age 2.
This early stage of language learning is all the more miraculous because it happens before the infant has even learned any words. Think about it. A 6-month-old wouldn't be able to use the knowledge that fit and fat are two different words, meaning two different things, to deduce the fact that "i" and "a" are two meaningfully different vowel sounds. So how does the baby do it?
Questions like this are one of the things that make the study of language acquisition so fascinating. Theories about how infants pull off the feat are things that linguists tend to disagree about, even when they do agree on the data. Here's one way it might work.
Any meaningful sound in a language's sound system — what linguists call a phoneme — is actually something of an idealized category containing a range of objectively slightly different sounds. Take the phoneme /k/, for example. In the word keep, that "k" sound is actually articulated a little further forward that in the word cool. Why? The sound following the "k" sound in keep is a front vowel, articulated with the tongue pushed towards the front of the mouth. The vowel sound in cool, on the other hand, is pronounced further back. Your tongue adjusts itself a little in anticipation of the following vowel — a process linguists call assimilation — so that the result is that the "k" in keep is actually articulated a little further forward than the "k" in cool.
For an English speaker, that slight difference in articulation is completely unimportant. That's why you probably can't even hear it. Your ear and your brain have learned to ignore it because a front "k" and a back "k" are not two meaningfully different sounds in your language. If you substituted a front for a back "k," it wouldn't change the meaning of a word. They're both members of the single phoneme /k/, with slight differences in articulation determined by incidental things such as the vowel that follows. You might say that a prototypical, or central, example of the "k" phoneme exerts a kind of "magnet effect" on slightly different variants, pulling them into the category such that, perceptually, all the minor variants count as "the same" even thought there are in fact slight objective acoustic differences between them.
Experiments by University of Washington language acquisition researcher Patricia Kuhl have demonstrated that, even at the age of 6 months, Swedish and American infants show evidence of different magnet effect patterns matching differences in the two languages’ vowel systems. An American 6-month-old will hear a range of high front vowel sounds — the almost-identical vowel sounds in words such as me, keep, and read — and consider them to be all "the same" as each other, and as a prototypical, central member of that phoneme category. A Swedish six-month-old, meanwhile, will hear the same high front vowel sounds and pull them into a quite different phoneme category, the high front rounded phoneme /y/. A central example of the Swedish /y/ phoneme would be pronounced with your tongue in the position of the English vowel in keep, but with the lips rounded as in a word such as moot. So the specific "magnet effect" patterns a 6-month-old will display depend on the phoneme system of the particular language the 6-month-old is getting ready to learn.
Even though any phoneme in any language will permit a certain amount of variation among its non-prototypical members — front and back "k" in English, for example — that variation is constrained by the other phonemes in the system. If the difference between a front and a back "k" were to change the meaning of a word, there would be an automatic constraint against letting the variants of the phoneme stray too far. The back "k" phoneme, in other words, would fail to exert a magnet effect on front "k" variants — perceptually, they would be kept distinct. That also means that a speaker of a language with two different "k" phonemes, one front and one back, would have an automatic constraint against letting the vowel following the "k" sound determine how far front or back it is. You can see that one result of this would be that front and back "k" would both appear in the same environment — with a high front vowel following either one, for example, rather than the vowel determining the pronunciation of the consonant. So if a 6-month-old infant is attuned to the distributional patterns of the sounds she hears in the language spoken around her — whether two acoustically similar consonant sounds do or don't appear in the same vowel environment, for example — then she can begin to solve the puzzle of which sound-differences are meaningful, and which are not.
If you were to think all this sounds like an incredibly difficult task, you'd be partly right. For an 18-year-old taking a college linguistics class, it's a real challenge to perform a distributional analysis on a body of linguistic data. But many aspects of language-learning that are hard for a grown-up or older child are much easier for an infant. The tools of analysis the infant is equipped with by nature, she later loses once that knowledge has done its job of helping her learn her native language. That's why infants put almost any grownup to shame in language-learning skill. And that's why high schools and colleges need to impose foreign language requirements, while parents couldn't prevent their child from learning their language even if they tried.
|
