DENVER - You may not know it, but you took a course in linguistics as a baby.
By listening to the talk around them, infants pick up sound patterns that help them understand the speech they hear, according to new research from the University of Wisconsin-Madison. But this research also shows that some patterns are easier to identify, suggesting that the development of human language may have been shaped by what infants could learn.
These results were presented here today, Monday, Feb. 17, at the annual meeting of the American Association for the Advancement of Science.
In a series of forthcoming papers, psychologist Jenny Saffran, who directs the Infant Learning Laboratory at UW-Madison, suggests how infants quickly acquire language, specifically their ability to find word boundaries - where words begin and end - from a steady stream of speech. "We've known for a long time that babies acquire language rapidly," she says, "but what we haven't known is how they do it."
In all her studies, Saffran introduces her infant listeners to an artificial, or nonsense, language. Examples of words include "giku," "tuka" and "bugo." By using these made-up words, which the tiny listeners have never heard before, Saffran can isolate particular elements found in natural languages such as English.
For just a couple of minutes, the infants hear dozens of two-syllable words strung together in a stream of monotone speech, unbroken by any pauses (for example, gikutukabugo...). The words are presented in a particular order that reveals a sound pattern. If babies recognize the pattern, says Saffran, they will use it to quickly identify word boundaries in what they hear next.
To test this, Saffran introduces her listeners to a new string of nonsense words in which only some of them fit the pattern heard earlier. Saffran records how long the infants listen to the parts that conform to the pattern and the parts that don't. A significant difference in times, she explains, means the infants did pick up the pattern.
As her recent studies show, infants do learn sound patterns, which then help them learn words and, ultimately, grammar. Their ability to do this, however, depends on age.
By exposing infants who are 6-and-a-half and 9 months old to a string of made-up words in a certain order, Saffran learned that the two age groups use different strategies to determine where words end and begin. While the younger listeners identified word boundaries by relying on the likelihood that certain sounds occur together, the older listeners paid attention to what speech sounds were emphasized, or stressed. Because 90 percent of two-syllable words in English follow the same stress pattern, says Saffran, infants can use the pattern to determine the word boundaries.
"At different points in development, babies orient towards some cues and not others," says Saffran. Why? "More linguistic experience." Before infants can recognize that stressed and unstressed syllables are reliable indicators of word boundaries, explains Saffran, they must first know a few words - lessons they learn earlier by learning which sounds are likely to occur together.
Findings from this study will be published in an upcoming issue of the journal Developmental Psychology.
Once infants go from syllables to words, they then can recognize simple grammars, according to Saffran's second study now in press at the journal Infancy. At age one year - just three months after babies begin using stress cues - infants can recognize patterns in word orderings. After listening to a continuous string of words in a particular order, the infants were able to identify permissible word orderings. Just as noted in the other study, Saffran says that only after prior learning can infants acquire additional language abilities: "Until they learn words, the grammar is invisible."
While these two studies looked at babies' ability to acquire sound patterns common in natural languages, a recent third study by the Wisconsin psychologist investigated infants' ability to acquire patterns not often heard in everyday speech. The question Saffran wanted to answer, she says, was, "'Does language work in a way that best fits the brain?'" In other words: Are certain sound patterns more common than others because they make it easier for infants to learn language? This study is in press at Developmental Psychology.
Unlike the other studies, which exposed infants to generalizations in language patterns, such as the grouping of sounds, this study tested an infant's ability to recognize something more specific - that syllables begin with some sounds, such as /p/, /d/ and /k/, but not others, such as /b/, /t/ and /g/. This pattern, says Saffran, is uncommon in phonological systems, which tend to place restrictions on types of sound segments, not individual ones.
As Saffran found when she measured how long the infants listened to words that did and didn't conform to the rare pattern, there was no significant difference in the listening times. This finding, she says, suggests that babies had difficulty acquiring the pattern.
The infants' difficulty in identifying the unusual sound pattern in this third study, she says, is likely to be the result of removing information helpful to young listeners as they acquire language. "There are certain types of patterns that they're better at picking up," adds Saffran. "Perhaps human languages have these patterns to make language more learnable. "
Asking questions about what an infant can't learn, she says, can be just as interesting and informative as asking ones about what they can learn. In addition to providing knowledge about language deficits in some children, the answers could offer clues to how human language first developed and how it has evolved.
NOTE TO PHOTO EDITORS/MULTIMEDIA EDITORS: To download high-resolution photos and a sample sound file to accompany this story, please visit: http://www.news.wisc.edu/newsphotos/saffran.html
- Emily Carlson 608-262-9772, firstname.lastname@example.org
AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert! system.