1998 LECTURE SERIES

Language, Brain and Mind: Early Experience Alters the Perception of Speech

Dr. Patricia Kuhl
Child Development Center
University of Washington
April 1, 1998

Language is arguably one of the most complex things that we engage in. We are now in a very unique time studying something like language in the brain. In my lifetime of research, about 25 years, I've been working on this issue and problem. I've never seen a case where so much interest, both by the academy and society, and even by business, is all focused on the same thing. Now why is this the case? Either from an academic perspective or from reading it in the press, one can conclude that this is the decade of the brain. We're trying to figure out how this device that we're carrying around works. We know that it's complicated. But recently there have been breakthroughs that make us understand that the brain is a physical, living thing, and we have to take care of it. We can help it develop, and we shouldn't think of it as just a device that's going to operate on it's own, without any care or feeding. It's like the rest of our bodies. And so speech and language provide the window on the brain. It helps us understand, in one small respect, how this complex device works. In universities across the country, there are many different people, computer scientists, engineers, linguists, psychologists, speech and hearing scientists, who are interested in the brain when it comes to this aspect of language and speech.

Are businessmen interested in language? Well, yes. And the reason for that is that within the next five years, every PC on the planet will have a microphone sitting right in front of it. And instead of us with our fingers keystroking to try to get computers to type our words or to do the things that we want it to do, we'll simply be using our voices to talk to it and the computer will unpack what it is that you're saying. So yes, business is interested in what we know about how language is processed by the brain. There are many companies throughout the country and in Europe and in Asia working on speech.

Data suggests that well before children develop their first words, which is right around one year old, they are listening to us talk, and their brains are mapping the information that they hear. Now most of you have heard a child pass through the milestones of language development. You know that at about three months, a baby will look at you and coo back and forth vowel-like sounds, and they seem intently interested at looking at your face and trying to do with their mouths what you're doing with yours. By seven months, you'll hear these babbled syllables. They sound like bah-bah-bah or mah-mah-mah. By about a year, you will hear first words, by 18 months, two-word combinations, and by three years of age, whether that child is growing up in Zimbabwe, Tel Aviv, Moscow, Paris, Seattle, or Irvine, those kids are going to be able to talk your leg off, using very complicated sentences and ideas. You can have a real conversation with a three-year-old.

So how does it work? What are we learning about it? I'm going to refer to the level of speech of consonants and vowels, the early perception of the building blocks that create words. We use speech sounds to test how this whole system gets wired, because we can study babies just hours old. We can watch them grow up in different countries and look at the results of nature's experiment. Nature organizes itself such that kids are growing up all over the world, listening, most of the time, to a single language. And when they're listening to that single language, we can take measure, every so often, and see how their beginning gets altered by listening to a particular language. The surprising news is that at these very early ages, there's evidence that just by simply listening to us talk, babies are developing a kind of mental map for speech that makes it ideal for them to acquire their own particular language. It sets them up for word learning.

We use to think words came first, and then they would sort out which sounds were important to code meaningful differences in their language. The research is now telling us that it is just the opposite. They work at the sound level first, understanding the sound patterns of the language, and understanding then promotes word acquisition when they reach one year of age.

Computers have trouble with speech because of the way we really talk. We say, didja hit it to Tom? And "didja" is not in the dictionary that the computer has. Now how do we do it? Well again, that's part of the puzzle. We don't have to hear, did you hit it to Tom? But every computer device on the market today has to have one second pauses between each word in order to interpret the speech. The thing to take from this is that we are going to be talking about these format frequencies of speech. I can generate those by computer and control them in very careful ways to do studies.

Now the studies I'm referring to are studies on babies that are very young, under one year of age. We have laboratories now in Japan, Russia, Stockholm, Finland, Germany, the United States and soon in France. We're watching these kids develop and testing how their perceptual systems change as a function of language experience.

When do we become listeners of a particular language rather than universal linguists? And the answer, again, is surprising. It's not as you might think, when we learn to read or when we've learn to talk in full blown language at age three or four. It's none of those ages. It's very early in development of approximately six months of age. It is in this very early period that a transition between a more universal approach to listening occurs. All across the world, we have seen in our studies that early in development this change occurs from an ability to hear all the distinctions to one that's more restricted.

We decided to start running babies in experiments in which we would vary the stimuli and ask the infants, with a head turn technique, which ones they could hear as different. When we decided to do this experiment, we had to think about what country to go to. We ended up in Stockholm. Within about four months, we tested 350 babies. We tested to determine whether or not the effects of language experience would alter the brains of infants at six months of age. At only six months old they don't understand words yet. They don't produce any words yet. They're just listening to sounds. Language is washing over them as they hear it, but it's doing more that. It's affecting their perceptual mechanism to allow them to hear a more unified category for the sounds of their language. They seemed to listen, but is it really having an impact on them? This study seems to suggest that it is.

It draws our attention to what we say to infants and what's going on in their brains when we say it. What do we know about this signal that we deliver to babies? We know there's a signal called "Motherese," or "Parentese," and now we've learned that even if you're not a mother or a father, and we put a baby in your arms, that you'll produce this special kind of speech. So sometimes you'll hear it refer to "Caretakerese." Now the speech that we produce to babies is very unique. So what are we doing when we talk to babies in this way? It sounds silly. We all think it sounds silly, and yet we know that babies prefer it. It has a much slower cadence, pauses between utterances, and it has an exaggerated and raised fundamental frequency.

Now in the laboratory, we can show that babies will choose this kind of speech, hands down, over language that we use when we talk to one another. So if you give a baby a kind of radio station test in the laboratory and they can listen to mothers, not their own mom, but mothers producing this kind of speech to another adult or this kind of speech, babies will do whatever they have to do to turn that signal on. They love listening to it. And the studies show that it does have to do with this pitch. The pitch contour. They like these pitch contours. We think that the pitch contour is a kind of acoustic hook. We get their attention to their speech patterns by using this type of speech. We all do it. It's instinctive. We don't know that we're doing it, but we do believe now that it is good for the infant's mapping of language. Not only because they like this pitch contour, but because of something we discovered. When we speak to our infants, what we do is make better vowels. We stretch them out. We make them long. Instead of saying, I bought a pair of beads, which is what we say to each other, we say, look at the beeeads. These are beeeads. Do you like these beeeads? Drag it out, over-articulate it, hyper-articulate it in fact. We think we do this to create better contrast for the baby, to show them how different EE, AH and OO are, and to allow them, give them many examples in this space, so that when they have to produce AH, EE, and OO with their little vocal tracks, they might know how to do it. We think that this production of unique Motherese or Parentese is good for infants. They like listening to it, and it provides a kind of tutorial on the units of speech.

Let's concentrate on one other aspect of language. One of the things that we do when we listen to speaker is watch their lips. We know that in noisy situations, and particularly for those of us who wear glasses, you think you hear better when you've got your glasses on. My husband said to me once, why are you putting your glasses on when you're on the telephone? And I said, I hear better when I've got my glasses on. He said, sure you do, dear. And I meant it. Meaning, that was superstitious behavior, but the truth is, you do. If you wear glasses, you will hear better with your glasses on because you're doing something while you're listening in watching the lips of the talker.

In an experiment with 18-week-old infants, we analyzed their eye movements to see where they look when people talk to them. Sure enough the 18-week-old infant is doing something akin to lip reading. They are beginning to map the correlation between the sight and the sound, and that's why as adults we show this auditory/visual illusion, because we can't really separate it. We've seen, that we can work with our eyes closed perfectly well. But if your eyes are open, processing that information is mandatory. You can't turn it off. If you see those lips doing something, your brain's going to process what sound that could be, given the lip movements. And so babies are starting that at a very, very early age.

When the babies are watching these faces, they will start to coo back to the face. Within about a minute, the babies will start to make their own little AHs and EEs and OOs. And the studies show that by 20 weeks of age, even though their vocal tracts are very small, they're producing sounds. This early mapping in memory that the baby is forming based on listening to a language, in this case English, is affecting their perceptual systems.

Early mental representation of the sounds that we're producing to babies is causing both their perceptual systems and their production systems to be altered by experience, and that promotes their language development.

Basically there's a three-step development that we're seeing in infants, where babies are born hearing many distinctions, many more than you would need for a single language. But by six months of age, we get the first evidence that listening to a language has altered their brain's processing of sound. The way we think it's being altered is by creating maps, that stretch and shrink the sound distinctions that you hear. Between categories, there's a stretching. Within a category, there's a kind of shrinking of the ability to hear the distinctions in sound, and that's very good for listening to a language.

Speech is kind of a multi-model event. Your brain pays attention to both to the auditory cues and the visual cues, and that's true even as a baby. You're paying attention both to the sight of the sound and the sound itself. In the future the tools of neuroscience will hopefully uncover all kinds of new things that we think will help explain something about language in normal individuals. We hope to be learning something about children with dyslexia, children with autism, children with Fragile X Syndrome. Each of these cases produces a different kind of language disorder. Many of these disorders rely on that early ability to handle the building blocks of speech, the phonetic units of speech. We hope that the work we're doing on normally developing infants across countries will tell us a great deal about things that might help us design interventions for kids who are not able to do that in their early years.