Science Watch

Babies who can master a simple language task when they're under a year old later do better on vocabulary and other language tests as toddlers and preschoolers, according to a study in the July Developmental Psychology (Vol. 42, No. 4). The research is among the first to suggest that infant language-learning tests can detect real individual differences in language ability. The researchers hope that the results may someday lead to earlier and simpler screenings for language impairment.

Until now, most researchers who studied infant language learning focused on group differences, such as how 7-month-olds differ from 8-month-olds, rather than individual differences. That's because it's difficult--if not impossible--to get a baby to sit down and pay attention all the time. One day Junior might be happy to stare at pictures or listen to words for the sake of science, but the next day he might be tired, teething or just in no mood to participate.

So if a particular baby does well--or poorly--in a particular experiment, researchers have reasoned, there's no way to know whether that performance indicates something intrinsic about the baby's abilities or just something fleeting about the baby's momentary state of mind.

"If, in a study, say 22 out of 24 7-month-olds could do a task, it's always been assumed that pretty much all infants that age could, because maybe the other two were just teething or cranky or had a wet diaper," explains study author Rochelle Newman, PhD, of the University of Maryland, College Park. "But what we were wondering was: What about the other infants? Were they just cranky that day, or did they really have poorer abilities?"

Old data, new study

To answer that question, Newman, University of Maryland psychologist Nan Ratner, PhD, and their colleagues reached into a trove of data collected by the late psychologist Peter Jusczyk, PhD, of Johns Hopkins University.

In 2000 and 2001, Jusczyk tested 412 7- to 12-month-old babies on their ability to pick out, or segment, individual words from a fluent stream of speech. This ability is crucial to learning and understanding a language. We all do it automatically in our native language, but when we listen to an unfamiliar language, we can't tell where one word ends and the next begins. Segmenting, then, is a necessary first step in language learning--for infants and adults.

In their experiment, Jusczyk and his colleagues repeated a single word, such as "cup," until the infants were familiar with it. Then, they played two recordings--one of a speech that included the word "cup" several times, and one of a speech that didn't. If the infants listened longer to the speech containing "cup," that was evidence that they were able to segment the words.

In their first experiment, the researchers reanalyzed Jusczyk's data from the perspective of individual differences. Jusczyk had followed the children in his studies for several years, so Newman also had access to data collected when the children were 2 years old. At that time, Juscyzk had asked the parents to fill out an inventory of their children's vocabulary. Newman decided to look at the data from the children who scored in the top 15 percent and the bottom 15 percent of the vocabulary inventory. She found that 71 percent of the high-scoring toddlers had successfully completed the language-segmenting task as infants, but only 38 percent of the low-scoring toddlers had done so.

Newman wanted to follow up to see how the children were doing by age 4 or 5. So in a second experiment, she and her colleagues tracked down 27 children--by now 4 to 6 years old--who had taken part in the original studies. Fourteen of the children had been "high" vocabulary scorers at age 2, and 13 had been "low" vocabulary scorers. The researchers gave the children a battery of tests, including an IQ test, a vocabulary assessment, a grammatical assessment, a test of articulation ability and other evaluations.

When they looked back at Juscyzk's original data, they found that the children's segmentation skill as babies predicted their language skills--vocabulary and grammar--at ages 4 to 6. However, it didn't predict their overall IQ, articulation ability or other skills.

Infants who can't segment words get off to a slower start, Newman says. "They're not impaired at age 5--they're in the range of normal--they're just not as good."

Earlier screening ahead?

The study reflects the infant-development field's new interest in individual differences and the predictive validity of infant language tests, according to developmental psychologist and infant researcher Anne Fernald, PhD, of Stanford University.

"I think this is an important question, and these are important findings," she says.

But she cautions that the study, which relies on retrospective archival data, needs to be backed up with a prospective, longitudinal study.

Developmental psychologist Patricia Kuhl, PhD, of the University of Washington, also praises the work. She's done research herself suggesting that some measures of phonetic perception taken at 6 months old can predict children's language skills more than two years later, when they are toddlers. "Newman's study is totally compatible with mine," she says. "Together they're showing us how infants crack the speech code."

Newman says that her next step will be to conduct the kind of longitudinal study that Fernald suggests. She plans to follow a cohort of children from infancy through early childhood, testing language skills along the way.

"Because we relied on archival data we weren't able to test all of the skills we would have liked. Segmentation is one, but there are others that are important," she says.

The researchers also plan to include both children who they expect to have normal language skills and those who are more likely to be impaired because of a family history of language impairments.

And if the researchers do find that some tasks do a good job of predicting which children will develop language impairments, Newman says, that could someday lead to earlier screening tests for those impairments.

But, she cautions, that day is a long way off. "Even if we do find ones that work, most of these are not easy paper-and-pencil tests. So we'd still have to find a way to develop ones that could be used easily in a doctor's office."