The Language Explosion

Acquiring a set of phonemes is a first step toward language, but just a baby step. To start decoding speech, you have to recognize words. And as anyone listening to a foreign conversation quickly discovers, people don't talk one... word... at... a... time. Real-life language-even the melodious "parentese" that parents use with infants--consists mainly of nonstop streams of sound. So how do babies suss out the boundaries? Long before they recognize words, says Peter Jusczyk, a cognitive scientist at Johns Hopkins University, they get a feel for how their language uses phonemes to launch syllables. By the time they're 7 months old, American babies are well accustomed to hearing t joined with r (as in tram) and c with l (as in clam), but they've been spared combinations like db, gd, kt, ts and ng, all of which occur in other languages. And once they have an ear for syllables, word boundaries become less mysterious. Ten / groaning / deadbeats / are / cleaning / a / train on / blacktop makes acoustic sense in English, even if you don't know the words. Te/ rgroanin / gdea / dbea / tsare / cleani / nga/ traino/ nbla / cktop isn't an option.

As children start to recognize and play with syllables, they also pick up on the metrical patterns among them. French words tend to end with a stressed syllable. The majority of English words-and virtually all of the mommy-daddy-baby-doggie diminutives that parents heap on children--have the accented syllable up front. Until they're 6 months old, American babies are no more responsive to words like bigger than they are to words like guitar. But Jusczyk has found that 6- to 10-month-olds develop a clear bias for words with first-syllable accents. They suck more vigorously when they hear such words, regardless of whether they're read from lists or tucked into streams of normal speech. The implication is that children less than a year old hear speech not as a blur of sound but as a series of distinct but meaningless words.

Meaning

BY THEIR FIRST BIRTHDAY, MOST KIDS start linking words to meanings. Amid their streams of sweet, melodic gibberish, they start to name things-ball, cup, bottle, doggie. And even those who don't speak for a while often gesture to show off their mastery of the nose, eyes, ears and toes. These may seem small steps; after all, most 1-year-olds are surrounded by people who insist on pointing and naming every object in sight. But as Pinker observes, making the right connections is a complicated business. How complicated? Imagine yourself surrounded by people speaking a strange language. A rabbit runs by, and someone shouts, "Gavagai!" What does the word mean? "Rabbit" may seem the obvious inference, but it's just one of countless logical alternatives. Gavagai could refer to that particular creature, or it could have a range of broader meanings, from "four-legged plant eater" to "furry thing in motion." How do kids get to the right level of generalization? Why don't they spend their lives trying to figure out what words like "rabbit" mean?

Because, says Stanford psychologist Ellen Markman, they come to the game with innate mental biases. Markman has shown that instead of testing endless hypotheses about each word's meaning, kids start from three basic assumptions. First, they figure that labels refer to whole objects, not parts or qualities. Second, they expect labels to denote classes of things (cups, balls, rabbits) rather than individual items. Third, they assume that anything with a name can have only one. These assumptions don't always lead directly to the right inference ("I'm not a noying," Dennis the Menace once told Mr. Wilson, "I'm a cowboy"). But they vastly simplify word learning. In keeping with the "whole object" assumption, a child won't consider a label for "handle" until she has one for "cup." And thanks to the "one label per object" assumption, a child who has mastered the word cup never assumes that handle is just another way of saying the same thing. "In that situation," says Mark-man, "the child accepts the possibility that the new word applies to some feature of the object."

Words accrue slowly at first. But around the age of 18 months, children's abilities explode. Most start acquiring new words at the phenomenal rate of one every two hours-and for the first time, they start combining them. Children don't all reach these milestones on exactly the same schedule; their development rates can vary by a year or more, and there's no evidence that late talkers end up less fluent than early talkers. But by their second birthdays, most kids have socked away 1,000 to 2,000 words and started tossing around two-word strings such as "no nap," "all wet" or "bottle juice."