Lending an Ear

John Trueswell and Lila Gleitman use adults to mimic the language-learning process in infants.
June 30, 2011

Parents, you might want to go easy on the flashcards. According to a new study conducted by John Trueswell, Professor of Psychology and Director of the Penn Institute for Research in Cognitive Science, and Lila Gleitman, Professor of Psychology and Linguistics, infants are more apt to learn their first words not from repeated viewings of visual representations, but instead during brief moments of clarity, usually in a dynamic, organic setting.

“There are two very different theories about how we go about learning words,” Trueswell says. “One theory is that children go through a process of visual tracking and tabulating across situations that eventually sorts out the ambiguities in the world. The model we’ve been working with, in contrast, is that a child can lock onto a word meaning almost instantaneously, based on a single observation of its use, and never forget it—a process called fast-mapping.”

Employing a sophisticated experimental procedure that Gleitman created, called the “Human Simulation Paradigm,” the team developed a method of evaluation that uses adults to simulate the role of the infants. Real mothers and fathers are videotaped going about their daily routine with their children—activities that are unscripted and organic.  The experimenters pick out instances of words that the parents say most often and make video clips of the 40 seconds or so during which these words occur. Adult study participants watch these videos with the sound turned off, only hearing a nonsense word, for example, “pilk” or “mipen,” just when the parent actually said the target word. The participants’ task is to guess the actual word the parent said, just by observing the videotape of the context.

“What we’ve found is that this ‘Human Simulation Paradigm,’ even though it uses adults, is currently the most naturalistic way to study what children observe when a word is uttered,” says Gleitman. “Rather than a flash card representing a language-learning scenario of the learner, participants are presented with a real-world, dynamic situation.”

“The model we’ve been working with is that a child can lock onto a word meaning almost instantaneously, based on a single observation of its use, and never forget it—a process called fast-mapping.” – John Trueswell

In order to stimulate the language-learning mind, the team chooses videos with varying degrees of ambiguity. Initially, participants might be shown what is referred to as a “high-informative video,” a scene in which the target word reference—“ball,” in this case—is pointed to by the parent as the nonsense word chimes in. From there they move on to low-informative scenes, scenarios that are much more ambiguous and might not even include the actual ball at all. This might involve a mother packing for a trip out to the playground and saying, “Let’s not forget the ball.”

In order to push the limits of different language-learning scenarios, Trueswell and Gleitman have begun to introduce strings of videos that alternate target words. In the first video, participants might be asked to identify “doll,” given only the word “mipen.” In the second, however, they might be asked to identify a new target word using a new nonsense word. In this scenario, participants are challenged to detect and store word references across multiple variations, mimicking an infant’s learning of new vocabulary in different situational contexts.

“What we’ve found is that participants who are introduced to high-informative videos early on are usually successful at identifying the target word in subsequent videos—a logical conclusion,” Trueswell says. “More importantly, however, is the fact that incorrect guesses seem to be very fleeting. Somehow the brain is able to quickly discard incorrect word references so that they don’t bog down the learning process, which speaks directly to the evolution of language learning in children.”

Trueswell and Gleitman hope to explore the language learning process even more in the future, possibly comparing population variables as well as differences in parental speech styles. As far as their inspiration for the project, that’s easy, says Gleitman, “I had the strange experience of having children,” she laughs. “As they began to acquire new language, it seemed like a miracle, so I rapidly shuffled my attention to learning everything I could about their experience.”