Going to the Source: Fisher Studies How Children Acquire Language

Cindy Fisher is a member of the Cognitive Science group whose experiments with young children power her research into how people learn their native language.

Researchers are famous for using discipline-specific jargon that leaves outsiders befuddled, but when Cindy Fisher uses terms like gorping and dacking even toddlers get it at least for her research purposes.

Fisher, a member of Beckman's Cognitive Science group and director of the Language Acquisition Laboratory at Illinois, studies how young children acquire their native language. In experiments with children as young as 15 months old, Fisher uses words that very young children might know from their own experiences, such as tickle, and made-up words like gorp to test theories about how they are learning syntax, meaning, and other linguistic features from the words they hear.

A recent article by Fisher and Beckman Fellow Yael Gertner published in the journal Psychological Science and titled Learning Words and Rules detailed the results of experiments looking at whether young children were able to abstractly represent elements like word order and use those abstract representations to generalize about other aspects of their native language.

"The looking time results of our experiments that we published with 25 and 21 month olds suggest that they do know something that general about what the order of words in a sentence means," said Fisher, a professor in the Developmental Division of the Department of Psychology. "So what we argued in our paper is that children represent their experiences with language in a format that includes abstract elements."

"We also are uncovering evidence that very young children represent their experience with language in this abstract format so that they are not necessarily tying everything to the meaning of particular words." - Cindy Fisher

Fisher said that means a child learning to speak English does know that a verb like tickle means something different than a verb like hit, but that "they also know that a tickler is an agent of action and the person being tickled is a recipient of action.

"So when they learn the significance of word order, for example, in a sentence containing tickle they go ah, in English the tickler is the one who comes first before the verb and the tickle-ee is the one who comes after the verb.

"If they also represent that as something like actor and acted upon then they can immediately generalize that to other verbs," she added. "So thats what we argued in that paper. The fact that children seem to generalize so quickly suggests that they do have those underlying abstract representations."

And those abstract representations, the article states, allow "children to rapidly detect general patterns in their native language, and thus to learn rules as well as words from the start."

The article in Psychological Science fits into what Fisher says is an ongoing debate in the field of language acquisition involving those who believe that an abstract language system is acquired quickly in a uniform and efficient manner, and those who think learning a language is messier, more piecemeal, and that children gain the ability to generalize about abstract syntactic and semantic elements in a more gradual way.

"Of course we think were striking a middle road; were not ignoring the lexical learning, thats a lot of what we study: can you learn that this verb is transitive, and what does this verb mean?" Fisher said. "But we also are uncovering evidence that very young children represent their experience with language in this abstract format so that they are not necessarily tying everything to the meaning of particular words."

In order to find that evidence, Fisher has to try and get into the mind of a toddler, something that isnt always easy to do.

"You have to make your experiment sort of self-explanatory and fun to get the little kids to sit through it," Fisher said. "Thats true when youre working with kids, it doesnt matter how little they are.

"Whereas grown-ups," she said with a laugh, "you pay them 10 bucks to do an experiment and they do what theyre told, within reason."

Fisher has spent the last 16 years trying to learn how young children acquire language, ever since she arrived at the University of Illinois following a Ph.D. at the University of Pennsylvania.

Fisher began researching language acquisition as a graduate student working with adults. She said one approach to this research topic is to do experiments with adults, computational studies, and simulate language acquisition using computers. Or you can, as Fisher puts it, go to the source.

"If youre thinking about sensitivity to particular cues, if we really want to know, well, do children represent their experience with language abstractly so that once they know that John tickled Mary means John was doing the tickling and Mary was being tickled that has consequences for sentences containing other verbs?" Fisher said. "There is no escape but to just find out in experiments with little kids."

Fisher said U of I Psychology Professor Rene Baillargeon was a guiding light when she first began designing experiments for children. She said designing those experiments is part intuition, part experience, and part trial and error.

"Imagine a little kid like a 15-month-old who can manage to understand a sentence with, say, a verb they already know," Fisher said. "If the verb is tickle they know exactly what tickle means, its a game they play with their mothers. They know what it means when they say John tickled Mary: that John was doing the tickling and Mary was being tickled.

"The question is when do children know that. When do they know that John tickled Mary and Mary tickled John mean different things? Its hard to tell from just the way they talk whether this is something they know about that particular verb or whether they know much more generally about how sentences go together in English."

Answering those questions is where Fishers background as an experimental psychologist comes in. But her projects arent limited to traditional experiments as a current collaboration with Dan Roth, a member of the Artificial Intelligence group with expertise in computational modeling, demonstrates.

Fishers work encompasses topics such as syntactic bootstrapping and structure-mapping, which she describes as "sort of a probabilistic, statistical version of knowing something about the syntax" of a sentence. Her research in that area showed that counting the nouns in a sentence can tell a lot about the meaning of a particular verb or an entire sentence. That research line led to a project with Roth that investigates how children mentally represent words in language learning and comprehension, examines the consequences of those representations in large samples of child-directed speech, and is working to a develop a semantic role labeling system toward possible future technology applications. The collaboration had a unique genesis.

"A long time ago I was giving a talk about this and Dan and I began to see that we had some very similar ideas about how you would learn to understand sentences," Fisher said. "He lent me a book by Dr. Seuss called Fox in Socks that had exactly this theory: ... Once you know the nouns, when those are combined in the sentence with some kind of relational word, wow, the relational word has got to be the relation between the fox and the socks. Thats basically what we think is happening."

Fisher said she enjoys working with children as young as 15 months because it adds an interesting element to the job of researcher and professor.

"Every line of work has a big range of what you spend your time doing, from big theorizing, figuring out how you are going to solve a particular problem, down to almost literally cut and paste. Before digital editing we did a lot more cut and paste than we do now," she said with a laugh. "So I like that mix."

Fisher said working with children means being creative about an experiments parameters and sometimes slowing things down.

"If kids of a certain age dont seem to understand the sentences in a particular experiment, they dont get it, there are always two possibilities: one is they really don't get it, they cant get it, they haven't learned whatever they need to learn about language in order to solve it," she said. "The other possibility is that you havent given them time to get it."

Even with adjustments, experiments involving young children dont always go as planned.

"All we try to do is make the experiment short and make the material engaging so that we can capture the kids attention," Fisher said. "But every now and then some kid is a movie critic and doesnt like our videos and wont watch. But that is OK, you cant please everybody."