Babies fish for words in a sea of chatter

Researcher studies how ‘little statisticians’ learn language

Strike up a conversation with psychologist Jenny Saffran, and you may encounter her peculiar second language. Just a few words, really. Words without meaning, but with an oddly alluring sound.

“Golabu … bidaku … daropi … pigola.”

This is the lexicon of Saffran’s research. From a string of nonsense words, uttered in a synthesized monotone to curious 8-month-olds, Saffran uncovers astonishing evidence of how infants assimilate their native language.

Saffran’s research has revealed a baby’s ability to pluck words from the dissonant chatter swirling around, and commit them to memory before they even learn to walk. They are “little statisticians,” in fact, capable of learning the statistical probability that a string of sounds represents a word.

“There’s an essential mystery to why language acquisition is so much easier for children than it is for adults,” she says. “How can infants do something that adults completely struggle with?”

Saffran, an assistant professor in UW–Madison’s psychology department, just finished her first year as a college professor. Her lab at the Waisman Center is still being set up. Yet at age 28, Saffran has already made a mark on the language development field and grabbed the attention of its biggest names.

Her fascination with language started when she was a high school student in Swarthmore, Pa. She learned she could receive academic credit for volunteering with a research project at Swarthmore College, and signed on with a professor who studied infant language development.

The experience made a permanent imprint, one that continually drew her back to language and linguistics studies. Infant language development became the topic of her Ph.D dissertation at the University of Rochester, and the target of her first major paper, which made the prestigious pages of the journal Science in 1996.

Having a paper published in Science as a graduate student was rare enough. But her paper managed to touch off an intellectual controversy.

Saffran’s research challenged a central dogma of the field, that language acquisition is a product not of learning, but of finely honed machinery in the brain that’s custom-built for acquiring language. In other words, the brain is already hard-wired with the capacity to fit language into a pre-existing structure.

But Saffran’s paper suggested there are equally sophisticated learning abilities at work in this process, and a child’s experiences in the first year of life play a much greater role than we imagined.

“Here I was a graduate student, and I was thrust into this high-level debate with some of the most prominent people in the field,” she says. “We were actually exchanging letters with Noam Chomsky.”

Chomsky, an MIT linguist considered the founding father of the field, was the first to propose that language acquisition is part of our biological makeup. “Having private letters from Chomsky was pretty intense, especially for a graduate student. I was a little starstruck,” she says.

“We knew the findings would be controversial, even though they shouldn’t be,” Saffran says of the nature vs. nurture arguments. “There’s a very fundamental clash. It’s almost like a religious debate.”

So what do “golabu” and “bidaku” have to do with these questions? They help Saffran show how infants can find word boundaries in a long stream of speech. Rather than individual words, Saffran has the babies listen to all of the words combined (“bidakupadotigolabu…”) over a two-minute period, with no pause or inflection between the syllables. The nonsense language is needed to guarantee the babies never encountered the words before.

To test their learning, she exposes babies to a repetition of one of the “words” in this string, and also to “nonwords” that contained syllables from the string, but in a different order. Consistently, the infants in the study listen longer to the “nonwords.”

The study takes advantage of a basic instinct with infants: What’s new is interesting, but what’s familiar becomes boring. What their findings told them is that the “nonwords” were novel to the infants, and kept their attention longer than did the “words” that they learned during the two-minute repetition.

By using different speakers and flashing lights to grab attention in her lab, Saffran can carefully observe when the babies’ attention is held and when it’s diverted. “Babies actually like it,” she says. “Kids love repetitive stuff, like rhymes and babble, and they are fascinated with flashing lights.”

In her new lab at the Waisman Center, Saffran can use the same arrangement to probe all sorts of related questions about the language mystery. Is this talent unique to language, or adaptable to other learning challenges? Do rhythmic regularities help children learn words? Can she measure an infant’s understanding of grammatical rules?

The million-dollar question, of course, is can we help babies become better language processors? Saffran thinks the answer is much simpler than some of the yuppie trends suggest, like teaching a 9-month-old to read. “Kids will learn language if you talk to them,” she says. “You don’t have to do anything special.”

In nature, Saffran finds an interesting parallel in the diverse songs of birds. Each species of bird has its distinct songs, like secret handshakes shared only with their species. It is so clinically precise, it clearly suggests a firm genetic imprint honed through centuries of evolution.

But laboratory studies have shown that chicks from some species cannot learn their call from a tape recording. And when paired with a bird of another species, they will not learn either song. It’s passed on only by parents of the same species.

In birds as in babies, the nature vs. nurture question will continue to fascinate Saffran. “It is still a mystery, which is wonderful for me,” she says. “We won’t know the answer to these things in my lifetime.”