Live from the Biggest Science Conference in the World: Baby Talk

Infants—to anyone other than their parents—can be a bore. Beyond cooing and crying, babies appear to be all sleep and bodily functions. But deep inside those cute, fuzzy little heads, infants are performing scores of staggering statistical feats. Bombarded with a bewildering range of sounds since birth, they possess mechanisms that scour these signals for statistical regularity, allowing them to emerge with something quite astonishing: an understanding of spoken language. As we age, this computational ability plummets. That's why it becomes so difficult to learn a second language, says Pat Kuhl of the University of Washington. Kuhl focused on the phonetics of language—the sounds we use to contrast words. In English, it's important to distinguish "ra" and "la" (as the statistical properties of our spoken language tell us), but in Mandarin this isn't the case, so Chinese people have a difficult time doing so. But Chinese babies—all babies, in fact—are what Kuhl calls "citizens of the world," able to distinguish between an enormous range of phonemes. It isn't until Chinese babies begin to learn the Mandarin language (about halfway through their first year) that their brains commit to the appropriate circuitry. This causes them to lose their ability to detect English-specific differences such as "ra" and "la," an ability that requires a different, "English" circuitry. Similarly, American babies can perceive Mandarin-specific differences when six months old, but begin to lose this ability towards the end of their first year. Kuhl wanted to see if she could take advantage of the baby brain's statistical abilities by exposing Americans to Mandarin, hoping they would retain more of their innate phonetic sensitivity. When the babies were nine months old, Kuhl began exposing them to Mandarin speakers three times a week for six weeks. The exposures were brief, but they made a big impact: When these American babies were a year old, they were still able to perceive all of the Mandarin-specific differences, performing as well as Taiwanese babies. An even more surprising find was the importance of social interaction. When Kuhl exposed the babies to an audio or video tape of Mandarin speakers for the same amount of time, the exposure had no impact--the babies did as poorly as their counterparts who had no exposure to Mandarin at all. This supports the idea that statistics plays a vital role in learning phonemes, and, importantly, suggests that the statistical input is gated by social interaction. Our social brains are modulating the way we glean data from the world.