How does an infant learn language? Once the basics of a language are learned, there are rules that someone familiar with language instinctively applies, but a baby coming to language with no previous knowledge cannot apply such rules. So how does it learn?
“It’s a big question," says Jessica Maye, an assistant professor of linguistics at Northwestern University in Evanston, Illinois, "because it takes money to make money. If you know something about language, you can use that to learn something else about language, but how do you learn the very first thing?”
That is one of the many questions about the human trait of language acquisition that Maye, who is affiliated with Northwestern’s Department of Communication Sciences and Disorders, has been pursuing since she began her undergraduate work at Seattle Pacific University in Washington in 1989.
A native of Northern California, Maye knew as early as high school that she wanted to become a linguist. It was a process of elimination: She enjoyed studying foreign languages, but she knew she wasn’t interested in becoming a translator, nor did she want to major in literature. She found both of those options too dry. However, she liked "the idea that in linguistics we approach the study of language as a science."
Once she had settled on linguistics, the decision to go to graduate school was, once again, a process of elimination. With her graduation from Seattle Pacific approaching, she asked her adviser what she could do with a bachelor’s degree in linguistics. "He said I could go back to school, or I could be a missionary and translate the bible into foreign languages. So I went to school," she says with a chuckle.
The Cognitive Approach
Early in her graduate school career at the University of Arizona, Tucson, Maye (pictured right) decided she wanted to focus on psycholinguistics, a relatively new branch of linguistics that draws on cognitive sciences, including psychology, computer science, artificial intelligence, speech and hearing, and neural imaging to explain how humans learn language.
After receiving her Ph.D. in 2000, Maye spent 3 years as a postdoctoral fellow in brain and cognitive sciences at the University of Rochester in New York, where she began conducting experiments on how babies learn. She used the period of her postdoc as a time for maturation. "I took some classes to help myself develop my own way of looking at things, to be more of an independent person than I had been in graduate school," she says. From Rochester, she moved to her current appointment at Northwestern in the Department of Communication Sciences and Disorders, where she has continued to study how infants learn language.
"Language is a very complex system of codes," says Maye, "a way we get meanings from our head into somebody else’s head. You learn it by hearing people and trying to figure out what they mean. It’s an example of how the human brain is able to make sense of really complex and highly variable and noisy input and attach it to categories to have meaning. What we would like to learn by studying language is how the human brain learns something that is that complicated without any direct evidence. You are not given the meaning of a sentence; you just hear it, and you have to guess what the meaning is."
Infants, Adults, and Monkeys
To shed light on precisely how such learning occurs, Maye, along with Dan Weiss, a psychologist at Pennsylvania State University in State College, initiated a study funded by the National Science Foundation (NSF) to look at differences in learning patterns in infants, adults, and monkeys. "It is exciting that each of these groups learn about speech sounds the same way, but what is really confusing is that only babies are ultimately successful at learning a new language really well," says Maye. "Adults do not learn a new language as well. They can learn to become very proficient, but they are going to be less proficient than someone who learns it as their first language. And nonhuman primates can learn to find the boundaries between words, but they don’t learn language at all."
The goal of the NSF-funded project is to determine why infants are so much more proficient than adults and monkeys. "If we know more about how infants use the statistical information they have access to, we can incorporate that information into machine processing of language to better enable speech recognition to correctly process language as humans do," says Matt Goldrick, an assistant professor in Northwestern’s Linguistics Department (which is separate from the Department of Communication Sciences and Disorders where Maye works). Maye's research, he says, "will give us insight into why certain individuals are having trouble learning language. Jessica’s research might provide insight into problems not in getting access to the information but in using it."
“Knowing how language learners make use of statistical information may suggest ways we can target language therapies," says Goldrick. Why “statistical?” "There is a pattern to the distribution of sounds and words in various contexts, and those patterns provide powerful cues to language structure."
Studies that commenced in the 1990s have indicated that language is learned in part by making use of these statistical regularities. So, many have wondered, what enables infants to make such good use of the information contained in the regularities, whereas adults, machines, and adult monkeys do much worse?
Past studies comparing the language-learning abilities of humans and other primates have looked at what monkeys can do with human language. “People then draw conclusions that the monkeys are not doing the same thing as humans because they are monkeys and not humans. Part of the problem with that conclusion is that they are testing monkeys on something that is not their own"--the monkeys'--"communication system,” explains Maye.
Maye attempts to solve this problem by turning the tables; she asks humans to learn monkey calls. The investigators then strive to determine whether human subjects show the same learning patterns for monkey calls as they do for human speech. Through such tests, "we will learn what the differences are between humans and monkeys based on the fact that we are humans and they are monkeys," says Maye. She hopes she'll then be able to determine "what differences are based on the fact that we had been testing them on a human system versus a monkey system."
But why study the language abilities of monkeys at all? Because insights into differences between adults and monkeys should help elucidate why infants are better language learners than adults. And this, in turn, should help Maye understand the special language-acquisition skills possessed by human infants.
People and Machines
Apart from the implications about language therapy, there's another practical side to Maye's research. Computer scientists aiming to improve computer speech recognition need to enhance the ability of machines to achieve “source separation,” the ability to distinguish and make sense of several voices talking at the same time even in the presence of background noise (although a phenomenon Maye has not researched herself). People are very good at this, especially when listening to their native language; current machines are much worse. If we knew what makes people good at isolating speech sources, it should be easier to teach machines to do the same.
Another thing humans can do very well but machines cannot is find the edges of words. Written words are separated by spaces that enable the reader to determine where one word ends and the next begins. But spoken language is a steady flow of sound with essentially no interruptions. Yet a listener familiar with a language has no trouble finding the boundaries between spoken words.
Statistical regularities appear to play an important role in the ability to recognize these boundaries. For example, developing sensitivity to the frequency with which one syllable follows another can help an infant learn to identify the edges of words. Using the word “pretty” as an example, Maye notes that "only a small number of syllables can come after 'pre' because it’s at the beginning of a word. The only syllables that can come next are from other words that start with 'pre' like pretend, precede, and predict. A much larger set of syllables can follow the syllable 'ty' because that’s at the end of a word. … So when you’re at a point where you can’t predict the next syllable very well, you’re likely to be at the end of a word."
A Broad Field
Maye’s field of psycholinguistics is just one of many branches of linguistics. Another subspecialty is phonology, which is Goldrick’s area of specialization. “I particularly focus on how people represent and process information for the purpose of producing speech,” he says. In English, for example, the sounds of the letters T and R can be used in consecutive order at the beginning of a word, as in the word try, but not in the opposite order.
The other facets of linguistics include sociolinguists, the study of the role languages play in the sociology of a culture; an example is how language is used to demonstrate social status. Another--morphology--involves the study of the parts of words--prefixes, roots, and suffixes--and how they combine to form words. Other linguists study semantics--how words are used to convey meaning, whereas still others study syntax--how words are put together to form sentences.
"The unifying theme of all the branches of linguistics," says Goldrick, "is trying to understand the structure and the patterns that exist in human languages.” This structure occurs on the level of basic sounds, through words, and on to sentences--and then there are the cultural variations on all these patterns. Some linguists cross over among specialties, and others take "a broad approach in terms of looking at the same aspect of language structure in a number of different ways," says Goldrick. "That is kind of what Jessica is doing. She is using comparative studies and studies of adult language acquisition and studies of infant language acquisition to all bear on the same question: How is it that we learn and represent a sound structure? I think it is a great example of some of the strongest work in linguistics today."
This material is based upon work supported by the National Science Foundation Grant No. SES-0549096. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s), and do not necessarily reflect the views of the National Science Foundation.
Comments, suggestions? Please send your feedback to our editor .
Victor D. Chase is a Yorktown Heights, New York-based writer. He can be contacted at firstname.lastname@example.org.
Photos courtesy of Northwestern University Infant Studies Center
Photos courtesy of Northwestern University Infant Studies Center