Monday, June 24, 2013
Quality Matters More Than Quantity for Word Learning
Several studies have shown that how much parents say to their children when they are very young is a good predictor of children’s vocabulary at the point when they begin school. In turn, a child’s vocabulary size at school entry strongly predicts level of success throughout schooling even into high school and college.
A new study by psychologists at the University of Pennsylvania now shows that early vocabulary improvement is likely to have more to do with the “quality” of the interactions in which the words are used rather than the sheer quantity of speech directed at young children. Moreover, the study shows that, unlike quantity, the quality of these interactions is not related to the parents’ socioeconomic status.
The study was conducted by professors John Trueswell and Lila Gleitman, both of the Department of Psychology in Penn’s School of Arts and Sciences, as well as by Erica Cartmill and Susan Goldin-Meadow of the University of Chicago. Also contributing to the study were Benjamin Armstrong III of Penn and Tamara Medina of Drexel University.
It was published in the Proceedings of the National Academy of Sciences.
Knowing how critical early-language acquisition is to a person’s future success, Trueswell and Gleitman have long investigated the mechanisms involved in how children learn their first words. One of their previous studies suggests that children learn these words in what might be described as a “eureka” moment — that is, only after “highly informative” examples of speech that clearly connect the word to the thing it refers to.
The researchers suspected these highly informative examples would matter much more than the sheer amount of talk in the home when it came to which children learned more words. To determine if this was the case, they set out to track the long-term effects of these examples, seeing if children who had been exposed to them more often did better on a vocabulary test three years later. However, to begin this study, the researchers first had to determine what constituted highly informative speech.
To quantify this phenomenon, the researchers visited more than 50 families from various backgrounds in their homes and videotaped parents interacting with their children. They made these visits when the children were 14 months old and then again four months later.
The researchers edited these taped interactions down to 40-second segments, each centered on one instance of a parent saying a common, concrete noun, such as “book,” “ball” or “dog.” The researchers showed these segments to adult volunteers but muted the video until the parent reached the target word, which was replaced by a beep. The volunteers were asked to guess the word the parent was saying in each instance.
“We purposely chose videos of parents interacting with their children in the home because of the complexity there,” Trueswell said. “Our intuitions are a little misleading; we think it’s going to be a simple environment, but there’s all sorts of things happening at once and changing on a second-by-second basis. Identifying a particular word’s referent, especially when you don’t know any words to begin with, is not a simple task.”
By taking out the verbal context, the volunteers experience the taped situations in the way the children experience it, as they don’t yet understand any of the words and must rely on environmental clues to first learn them. The researchers also discarded any examples where the child might already know the word in question. In those cases, volunteers might be able to pick up clues from the child rather than the parent, and the parent might be less conscientious about connecting the meaning of a word to its referent.
“We see that the more an environment maximizes the ‘here and nowness’ of speech, such as when a parent is gesturing or looking at the object in question, the more likely it is that an interaction will be highly informative,” Gleitman said. “And it turns out this is surprisingly hard to do; only 7 percent of the examples were able to be guessed correctly by more than half of the adults we showed them to.”
If more than half the adults could guess an example’s target word correctly, that suggested the interaction was highly informative. The researchers used this approach to determine approximately how frequently each child in the study heard these highly informative examples. They found a surprising amount of variability: the parents who provided the highest rate of highly informative examples did so 38 percent of the time, while those who provided the lowest rate did so only 4 percent of the time.
“This means that some parents are providing 10 times as much highly informative learning instances as others,” Gleitman said.
The effect of this discrepancy was clear when the researchers tracked how well each of the children did on a standard vocabulary test three years later. The more frequently a child heard highly informative examples of speech, the better he or she did on these tests.
Increasing the quantity of speech was also beneficial but only because it increased the number of chances parents had to provide highly informative examples.
“Fortunately, low-informative instances seem to be ignored,” Trueswell said. “By talking to children more, it’s not as if you’re giving them bad data, you’re only increasing the opportunity to find those nuggets.”
Critically, the rate at which a parent gave highly informative examples to their children wasn’t correlated to the amount they spoke to them in total. This is potentially hopeful news, given the studies that link low socioeconomic status, or SES, to low speech quantity and thus to poor scholastic performance.
“There are a variety of reasons why low-SES parents are speaking less to their children,” Trueswell said, “but, when they do speak to them, their natural predispositions about talking about the ‘here and now’ don’t seem to be correlated to their SES.”
And while the exact mechanisms that lead to a particular bit of speech being highly informative will need to be determined in future research, the Penn team’s study shows how these quality examples can have an overriding and lasting effect on an important stage of a child’s development.
“You can see this effect even with all the variations in their lives and personalities,” Gleitman said. “Through all of that noise, the signal of a linear relationship between these highly informative examples and their children’s performance on that vocabulary test three years later shines through.”