By the time babies start talking, they probably have a good idea what the words they’re using mean. How that happens remains something of a mystery a lab at Duke University is trying to unravel.
Psychology professor Elika Bergelson’s team published a paper just before Thanksgiving that says it’s looking more and more likely that by the time they’re 6 months old, babies are associating specific words with specific objects.
And they’re not easily fooled when someone says, for example, the word “juice” while showing them a picture of a glass of juice and a picture of a car at the same time, Bergelson said.
Analyzing their eye movements, it’s easy to tell that “they track to the right picture,” said Bergelson, who joined Duke’s faculty in 2016 but started the project while she was at the University of Rochester. “They did a good job.”
Never miss a local story.
That still begs the question of how a baby makes the connection, but the study suggests it helps when mothers and other older people in the room use the words that describe the things the infant is interested in.
Seemingly, “when parents take the time to zero in and indicate what they’re talking about, that makes for a more teachable moment,” Bergelson said.
The paper, published Nov. 20 in the Proceedings of the National Academy of Sciences, built on a 2012 study she also worked on that showed by 6 to 9 months babies grasp the meaning of common words.
The latest report summarized the work Bergelson, former University of Rochester professor Richard Aslin and other members of their research team did with a group of 51 infants from the Rochester area.
After convincing families to participate, the team gathered two sets of observations.
One came from having parents bring their children to the lab so that researchers could closely track the way the babies’ eyes moved as they looked at a sequence of pictures the team always showed them two at a time. Sometimes the pictures were completely unrelated, like the aforementioned glass of juice and a car. Sometimes, the pairs were related, a stroller and a car being one selection there.
“All we do is put a little sticker on their forehead” and let the baby sit in its parent’s lap, Bergelson said. A camera tracks where the baby is looking.
The other involved sending people to the babies’ homes to collect an hour’s worth of video and a day’s worth of audio. With the recordings in hand, a platoon of research assistants went to work on a “brute-force annotation” to document what things were in a room, the words (nouns specifically) the people in themwere using and where the babies seemed to be looking.
Babies being babies, the process wasn’t necessarily a smooth one, especially with the in-lab eye-tracking measurements. While many of the tots seemed to enjoy the process, the paper noted the team had to toss some results because of “infant fussiness.”
“Sometimes they just decide, ‘I don’t want to sit here and look at a screen,’ ‘I’d like to feed,’ ‘I’d like my diaper changed,’ or ‘I’m over it,’” Bergelson said.
And given that the researchers were working with a small group, they also discarded the measurements from one infant whose word-association abilities were actually too good, lest the numbers from an outlier skew the overall statistical findings.
Not solved yet
The professor cautions that she and other researchers are still a long way from solving the mystery.
In the lab, babies generally didn’t seem to respond as accurately to a word when they were seeing pictures of related objects, for example the stroller and the car. It’s not clear whether that’s because they know enough about the resemblance to be confused, or whether they’re confused because they don’t understand there’s a resemblance in the first place.
Either way, the results are “consistent with babies having a notion that some things are more related than others,” Bergelson said.
And it confirms prior research and “parents’ intuition that talking to your kid helps and talking about the things they’re paying attention to helps” in building their vocabulary, she said.
The team remains in touch with the families that volunteered for the study, and is now getting ready to do some follow-up work now that the children are toddlers.
“It would be exciting to find that something we found at 6 or 12 months actually predicted kindergarten reading scores, but that’s getting way ahead of ourselves,” Bergelson said.
The work also needs to expand to take in families beyond the mostly middle- and upper-income ones who’ve participated so far.
“It’s good to get the lay of the land here, but it’s important to see how this works in other home environments,” Bergelson said.