Parents and teachers often use flashcards and picture books to teach young children new words, but a new study suggests that understanding basic words may come from a flash of initial insight more than repetition.
“What we know is children are getting a lot of input from their world, and they are teasing out what information is useful or not useful,” said Janice H. Im, the interim chief program officer for the Washington-based nonprofit Zero to Three: the National Center for Infants, Toddlers and Families. “If language experiences are not rich, then where is your interest to retain them?”
The study’s findings suggest that children—and, in fact, all new language learners—can build up concrete vocabulary from interacting with a complex learning environment, not just repeated exposure to words in isolation.
In a study published in the May issue of the Proceedings of the National Academy of Sciences, researchers at the University of Pennsylvania and Harvard University conducted a series of four experiments on how adults and preschool children learn the meaning of unfamiliar words. The researchers focused on so-called “seed words,” basic nouns that form the foundation for text comprehension.
Repetition vs. Insight
Many language-development researchers believe children learn a new word gradually, taking a general meaning from encountering it multiple times in various contexts and gradually arriving at a more specific meaning. By contrast, the researchers for the new study argue that people instead make a best guess about a new word’s meaning based on the context in which they initially encounter it, and hold onto the meaning unless it is clearly found to be wrong.
“Where people were learning gradually, they were learning the wrong thing. They got more and more abstract descriptions in order to cover all the examples,” said Lila R. Gleitman, a study co-author and a psychology professor at the University of Pennsylvania in Philadelphia. “But little children don’t learn that way at all; they learn concrete before the abstract; they learn doll before they learn toy.”
Those learning a new language at any age tend to follow that same early process, she said. In three experiments, including 37 adults and a replication study with a dozen 3- to 5-year-olds, participants watched short video scenes of a mother talking to her toddler in natural environments, like a playroom or a kitchen. The videos were muted to replicate the experience of someone at the very start of language learning, with a single word replaced by either a beep or a nonsense word, to evoke the experience of hearing a new word for the first time. Participants were asked to identify the meaning of the target word from context, which varied from one scene to the next. For a target word meaning “horse,” for example, the “parent” in the film might point to a toy horse and name it directly, or refer to it by saying, “Let’s see the horses today.”
For 90 percent of the scenes, no more than 30 percent of participants accurately identified the word. Of the 7 percent of scenes in which a majority of participants identified the mystery word, all of them were for concrete nouns naming basic groups of objects, such as “ball” and “horse,” which often are among the first gained in a child’s vocabulary. The scenes in which a majority of participants identified the mystery word were considered “highly informative.”
Children mirrored the identification pattern of the adults; they identified the target word in 53 percent of the “highly informative” scenes using common nouns, compared with only 22 percent of the other scenes.
Next, the researchers allowed the participants to try to identify a dozen new mystery words by viewing five scenes for each word; each scene used the word in a different context and across different word orders. If a child learns basic words through association, statistically comparing possible meanings over time, the researchers expected all the participants to improve steadily in their ability to guess the mystery words.
That’s not what happened. Instead, participants seemed to make a best initial guess at what the word meant and changed their minds only if the meaning was clearly wrong in a later scene. If they had a highly informative scene early on, 66 percent were able to identify the target word correctly, but their accuracy decreased, rather than increased, after watching the five scenes. Participants who saw the more-informative scenes later in the lineup were less likely to correct earlier misconceptions of the words.
Penn’s Ms. Gleitman suggested that people may mentally disregard examples that don’t fit a preconceived idea of a word’s meaning. “How can you avoid going abstract? Only by forgetting what you learned before,” she said. “What you remember is your guess, your one guess. It’s the failure of memory that’s rescuing the learning procedure—because you don’t remember the things that are wrong. It’s paradoxical, but that’s what seems to be happening.”
Moreover, she said, that could explain why educators do not see young children make large numbers of mistakes about the definitions of their first words, even during the first five years when they learn 5,000 to 6,000 words.
Ms. Im of Zero to Three said she has seen similar behavior with children she has worked with as well as her own daughter; children understand the meaning of a specific noun like “ball” before the more abstract “object.” “Why we call something ‘furniture’ or‘chair’ is really arbitrary,” Ms. Im said. “When children see that, they are able to grasp not only the meaning of a word but how it is used in a particular context.”
Fast-Mapping
Bob McMurray, an assistant psychology professor and the director of the Mechanisms of Audio-visual Categorization Lab at the University of Iowa in Iowa City, argues that both the “first, best guess” model known as fast-mapping and associative learning are likely at play in early language development. In a 2008 study published in Infancy, the journal of the International Society on Infant Studies, University of Iowa researchers showed 16 2-year-olds two familiar toys shaped like a car and a duck, and one new toy. When the toddlers were told, “Get the blicket!” more than 80 percent were able to retrieve the new toy and remember its name. However, after a five-minute delay, the children were not able to name the new object in a group of unfamiliar toys. “This research has demonstrated that learning does occur during a fast-mapping trial; however, the amount of learning from a single fast-mapping trial is insufficient to support full-blown word learning,” Mr. McMurray wrote in an essay on the project.
Yet Ms. Gleitman’s research may help explain the results of Deb Roy’s Human Speechome Project at the Massachusetts Institute of Technology. Mr. Roy recorded and mapped more than 230,000 hours of video in his son’s first three years. Mr. Roy, now the chief executive officer of Bluefin Labs in Cambridge, Mass., found his son’s first words were associated with “hot spots” in the home where the words often were used by adults, such as “water” in the kitchen.
