University Place

cc >> Welcome Everybody to Wednesday Nite at the Lab. I'm Tom Zinnen. I work here at the University of Wisconsin-Madison Biotechnology Center. I also work for UW-Extension Cooperative Extension. And on behalf of the Wisconsin Alumni Association, I would like to welcome you to Wednesday Nite at the Lab. We do Wednesday Nite at the Lab every Wednesday night, 50 times a year. Tonight, I'm delighted to be able to introduce to you Dr. Jenny Saffran who has been on the faculty of UW-Madison since 1997, which was the year that she received her PhD in Brain and Cognitive Sciences, and Linguistics, from the University of Rochester. She is currently the College of Letters and Science's Distinguished Professor of Psychology, and is an investigator at the Waisman Center. Her lab which is called the Infant Learning Lab has an international reputation in the area of Developmental Psychology. She has received a number of awards, including several early career awards from the American Psychological Association. And she received the Presidential Early Career Award for Scientists and Engineers from President Clinton. Dr. Saffran has also received several teaching awards on the UW-Madison Campus. She is married to fellow Psychology Professor Dr. Seth Pollack, and is the proud mom of two kids who are still in the throes of learning language. Would you please join me in welcoming Dr. Saffran to Wednesday Nite at the Lab. ( applause ) >> Okay are we live? Are we good? Okay. Excellent. I apologize that I didn't bring any hands-on exhibits today of babies, but I do want to start by saying that my lab is just on the other end of campus at the Waisman Center. We test a lot of babies, so thanks to Claire and her brother, and the other, oh, 1,198 babies we tested last year. And our goal is to begin to try to understand how babies learn language. I won't pretend to say that we know the answer. I feel like this is something that people will still be studying long after I'm retired and gone. And that's part of why I think it's such a great question for scientists young and old to tackle. So, visitors are always welcome, even if you don't have a baby to bring. We'd still love for you to come. So, I'd like to start with my conclusions just in case any of you have to leave early. My conclusion is really very simple, that babies are extraordinary learners. And that shouldn't surprise any of us who have ever hung out with babies. It's pretty obvious they're just soaking up knowledge every which way. And what I want to try to do tonight is to try to give you a bit of a better flavor of how some of that soaking up of knowledge is going on. What kinds of things babies are able to learn and how. And to try to draw some broader implications from this work, as well, for things like what kinds of environments we should structure for infants, or should not structure for infants given the types of ways they like to learn. So, for those of you who are parents, grandparents, caregivers, teachers, other people who are interested in how learning actually works in the real world, you can draw your own conclusions from what I have to say about what types of environments might be best for babies. So recently, the media and toy companies, and all sorts of other types of elements have caught on to something that parents have really known forever, that babies are smart. And they're smart in lots of neat ways. And every year, the major news magazines have a splashy cover about all the brilliant ways that babies are smart. And this goes really far in our current culture, because babies are having experiences. It's amazing what happens if you go to Google Images, and just type in things like "Baby and iPod," and you get scary stuff. You can actually buy that little Onesie there, that little iPod cover. Babies are having experiences that we did not evolve for. We did not evolve to watch DVDs or listen to iPods. And babies are experiencing things in ways that are unimaginable even a generation or two ago. So just to set the stage, just for a moment, I'd like you guys, and Claire can do a really good job of this for me, just stop being your educated adult selves and try to imagine that you're a baby. So just stop for a moment and take stock of your environment. You're a baby. You don't know what the words I'm saying mean. You don't even know that I'm saying words. There are these things that are touching you, and there's all this stuff all around you, and these weird sounds and smells and none of it makes any sense at the outset of learning. The point here is that the baby's environment, the typical baby's environment that we evolved in is wildly stimulated. I mean, if you just look around any environment you could imagine, it's wildly stimulating. Now of course some environments are more stimulating than others. ( laughter ) That's not my child. Again, my undergraduates delight in sending me good pictures for these sorts of purposes. That's one, here's another. Something I'm sure that Claire doesn't do to her little brother. ( laughter ) So, these are funny, but there is a serious point here, which is that the kid's environments are extremely rich and stimulating. Now some of you might have you know seen some of the hub-bub recently about things like "Baby Einstein," and is it good, is it bad. Should we be giving our babies all this extra stimulation? I'm not going to weigh in on this other than to say that of course--. This is another great one. If you just go to Google News and put in "Baby Einstein" you find all sorts of good stuff. The truth, of course, is that everything's fine in moderation. I'm not going to weigh in strongly here. But, did we evolve to do this? Is this what our brains are meant to be learning from, is this the manner in which we're meant to learn? What was that? ( inaudible ) Like I said, I'm not going to weigh on this here, but if people are particularly interested in this issue, there's a very good book out there called "Einstein Never Used Flash Cards," which really makes a scientific case against some of these toys and educational devices that are permeating our environment. So that said, if our environments are rich, how do we learn from them? How do we gather knowledge in this incredibly rich--. This also is not from my lab--. in this incredibly rich environment. And so that's what I'm going to talk about today. There are two forms of learning in infancy that we can sort of distinguish between. I'll tell you about both of them, and then I'm going to focus for almost all the rest of the talk on the second one. The first one is the more obvious one. It's learning by doing. Learning by exploring. Those of you who are familiar with Piaget, these are a lot of the ideas that came out of Piaget's work. The idea that part of how you learn is by pointing, and saying, "What's that?" or by locomoting across a space to get to what you want. That's the sort of obvious thing that we think about when we think about little kids, babies and toddlers learning. And that certainly is important. I'm going to give you a really neat example of that in a moment from someone else's research. What has become very interesting of late is something that we actually don't think about, something that's actually going on all the time, in you and me, and babies. But it's the most fascinating in babies. And that's what I call passive learning, learning by soaking up the structure that lives in our environment. And I'm going to show you today some examples from research in my own lab that suggest that babies are incredibly in tune to patterns in their environment. And they soak them up without being instructed to, without having them pointed out, and without being rewarded for soaking them up. They're learning all the time. So are we. So are we, but they're really extraordinary at it. So I'm just noting here, when I say "learning," I don't mean learning by teaching in the very explicit way we see in the educational setting, or by parents explicitly training their children or teaching their children. I mean learning either by doing, or learning by soaking up things. So here's an example. This is not a language example, but this is one of my favorite examples of active learning. It's the only one I'm going to talk about. This is an example with three-month old infants. And here's something to know about three-month-olds, two things to know. One is that they know very little about objects. So if you do experiments where you show them images of objects intersecting with each other, objects going behind each other, objects covering each other up, they don't really know what to make of it. They don't really understand that objects exist separate from one another, and that objects continue on the trajectory in space. They don't have a very good grasp of physics. The other thing to know about three-month-olds is that they're terrible at reaching for stuff. So if you've ever watched a three-month-old, let's say they want their pacifier or something like that. They can't do it, they don't know the motor control to do it. So what this researcher did was something that I think is just so ingenious. I'm amazed that no toy companies have yet forced her to market this, which is she equipped babies with little Velcro mittens and she gave the babies toys with Velcro on them. And so what happens, I've seen videos on this, is that the babies sort of bat around, because they don't know what they're doing, and inadvertently they hit the toy. And because of the Velcro, they can pick it up. Fascinating. And they explore, because babies really are little scientists in the same ways that we are. They like to test hypotheses. And they explore their environment. And so, they learn to play with these Velcro gloves. And it turns out that after a little bit of this mitten training, then when you test them again, they know a lot more about objects than babies of the same age who haven't had the mitten training. So that would be an example of active learning. Learning by acting on something and discovering something about your environment that you wouldn't know otherwise. Very Cool. And I'm just plugging a few books written by colleagues. This book is actually the book when friends of mine have babies, this is the book that I recommend more than any other. It's a book really about these sorts of active scientific explorations by young infants. So that's cool. That's not what I do. I do something different. This is not from my lab either, but it is kind of a cool picture. I study what babies can learn without any sort of active engagement, without trying, at least in any ways that we can see. And so, in my work I focus on language for two reasons. Well, the third reason is that languages are really interesting and they're cool. The first reason is that language is arguably the most complicated thing any of us will ever have to learn. I mean, we know this is complicated, because linguists can spend decades trying to describe, say, the grammar of a particular language and still not get it quite right. It's probably the most complicated human invention of all time. The most complicated cultural artifact of all time. So that's part of what makes learning it interesting. The second thing that makes its learning interesting is that it's done so quickly and so young. By the time kids are, they say three years of age, if you test them the right way, you can show that they know almost everything there is to know about how their language works. It doesn't come out when they talk, because it's hard to talk. But if you test what they know, they're almost done by the time they're three, except for learning more vocabulary. And that's extraordinary because this is hard stuff. And at the end of my talk I'll come back to why it seems to be so much harder for me, as an adult, than it is for the average two year old. So this is why I think language is so interesting. So, let me just give you a little bit of background on what babies know about language in the first year. And again, part of why this is so interesting is that they know a lot before you can really tell that they know it. So on average, just going by the milestones, babies will produce their first spoken word at about 12 months of age. Their "mama" or their "baba," or whatever. The caveat here is that it's actually really hard to know what counts as a word. If you've ever tried to figure out whether it's really a baby's first word, or they're just babbling, it's pretty tricky. But, on average, it's around 12 months. I have in parenthesis here, "earlier for signs." This is extremely interesting. It turns out that infants who are exposed to sign languages, either naturally say by having a parent who speaks American Sign Language, or by design like being taught a few signs like "more." The first signs emerge a couple months before the first words, spoken words, suggesting that it actually really is easier to do this "more" than it is to say more. So, we're waiting until like 12 months before we see strong evidence of having learned much about language, if we wait for what they say or sign. However, what they're saying or signing vastly underestimates what they know about their native language. And in the past decade, researchers in my field have done an amazing amount of work to try to figure out what babies know and when. And it's actually pretty cool. Here are just a few things, sort of fun facts about babies and language. One is that at birth, or as early as you can test infants, they can discriminate between the sounds of all the languages of the world. It doesn't matter what language you're going to learn, you're born able to perceive the sounds of all possible languages. So when we were born, even if we grew up learning English before a certain age, we could perceive all the sounds of languages other than English. I'll come back to that in a moment. Also from birth, infants prefer voices over pretty much any stimulus you can create. So, if you give them a choice, and the way people would do this, there are a couple different ways. One way you can give babies a choice is you attach a pacifier to a speaker and the pacifier will turn the speaker on and off, and the harder the baby sucks on the pacifier, the more they get to hear a particular sound. And they will suck harder to hear speech than to hear pretty much anything else. And interestingly, even within speech, they will suck harder or turn their head faster, or longer, to hear what we call infant directed speech, the un-PC term for this is "Motherese." It used to be called "Mmotherese," but that's not PC, because dads do it too. And we do it to dogs, and so that's kind of interesting.

But this is the way of speaking

( high-pitched and sing-songy ) Look at the turtle. Do you see mommy's turtle? I love your turtle. ( normal voice ) Parents the world over do this. This is not just an artifact of being uptight, upper-middle class Americans. This happens around the world. And it's actually one of the first mechanisms by which infants get both some linguistic information, but also emotional information. Because some of the ways we do this around the world are the same. So if you look at how parents talk around the world, sort of short sharp sounds, "uh-uh-uh." Those are sort of prohibitions around the world. "No!" "Don't do it!" Whereas, soothing "oh, ooh" are sort of soothing sounds around the world. We do this in music, too. It turns out there is infant directed song. It's very interesting. Babies prefer it. So that's neat. Now what does all this attention buy the baby? Well, it buys the baby a lot. By four months of age, and for those of you who haven't hung out with babies recently, four-month-olds are picking up their heads. But they're not rolling over typically, or maybe they're just starting to roll over. I mean, they're not doing much that you can see. But it turns out that they recognize their own names. Now, how do we know that? And this is where I'm going to start to give you guys a bit of a primer on how we do research with babies, because they can't talk, right. So, how do we get into their heads? When I say they prefer X or Y, or they know X or Y, what makes me think we know it? What makes me think they know it? Well, here is how the study was done. Okay, here's a little apparatus. Let's see, I think there's a pointer here. Actually, I'll just use this, you guys can see that right? So, here is a caregiver, and here is her infant seated on her lap. And the way this particular kind of study works, is on one side of the room on some trials, the babies will hear one name. So let's say I'm Jenny, and some other baby's name is Susan.

On half the trials they'll hear

( in different tones ) "Jenny, Jenny, Jenny, Jenny." And it will play as long as the baby keeps looking in the direction of that sound. When the baby looks away, the sound gets turned off. And the computer times how long the baby listened. On other trials, the baby might

hear

( in different tones ) "Susan, Susan, Susan, Susan." And it turns out that for a baby named Jenny, that baby will listen longer when they're hearing "Jenny" than when they're hearing "Susan." But for a baby whose name is Susan, they'll listen longer to "Susan" than to "Jenny." Now, does this mean that the baby knows that Jenny refers to her, is her name? No. It doesn't mean the babies know "Oh, I'm Jenny, and you're Susan." They don't know that kind of stuff. What it means so is that something they've heard really frequently in their environment, like their name, has become familiar to them. And they will recognize it in one of these kinds of experiments. And I think that the fact that a four-month-old, a little squishy critter that can barely seem to do much other than, you know, nurse and fill diapers and smile. And they've learned their name. Not that it's their name, but they've learned that certain sounds are more familiar than others. And I think that's really extraordinary. Around the same time, they start to winnow out the sounds of their language. Remember how I said a few minutes ago that babies begin as citizens of the world. They recognize the sounds of all the languages of the world that anyone has every tested. Well, sooner or later they stop recognizing all the sounds of all the languages of the world. And what's really extraordinary about this is how young this happens. I'm going to do this without hurting myself. There you go. So by the time babies are six months of age, again, still squishy, not really talking, maybe babbling or cooing a little bit. They can discriminate only the vowels of their native language. And by the time they're ten months of age, they can discriminate only the consonants of their native language. So, just to make this a little clearer, let me give you an example. So, in Japanese and Korean, the sounds "Ra" and "La," they're the same sounds. So, if you had words like rate and late in Japanese or Korean, they would be the same word. You can't tell the difference between them. It turns out that by the time Japanese and Korean babies are ten months of age, if you play "Ra, Ra, Ra, Ra" and switch to "La, La, La," they don't notice. They don't hear the switch. But a ten-month-old English learning infant, or French learning infant, would hear that switch. The Japanese babies could hear the switch a couple of months earlier, but by the time they're ten months of age, they've learned to ignore the difference between "R" and "L." Now, lest you think that--. English speakers always seem to think that they're the center of the world. But there's actually a lot of sounds that we in English can't perceive, either. There's no one here who speaks Hindi by any chance is there? You never know. Okay, I'll try to do it myself. In Hindi, there is a discrimination between "dah" and "dthah." ( both sounded the same ) Hear the difference? I can't hear it either. But I happen to know how those sounds are made. In Hindi, "dah" would mean one thing and "dthah" would mean a totally different thing. But in English they don't. When you guys were six months old, you could hear the difference between those sounds. By the time you were ten months old, you could no longer hear the difference between those sounds. And this is amazing. I mean they're learning to focus in on the sounds that matter in their native language and to ignore the ones that don't. Without being reinforced, without being told what to pay attention to, they're just doing it. And I think that's really cool. So how does this learning unfold? This is all fine and good. It's all very descriptive. I can tell you what babies know at different ages. This is my kid, so I can take credit there. ( light laughter ) How do they learn this stuff? In our research, what we've focused on is a somewhat counter intuitive theory of how babies learn. And at some level, it's counter intuitive. At another level, it makes perfect sense given what we know about brains. And what we think is going on, as part of what babies learn, and I'll show you some evidence for this, is that babies are little statisticians. So babies actually are tracking the statistics of their environment. They're actually keeping track of the probabilities with which events co-occur, the frequency with which events co-occur. And they're able to learn by essentially doing applied mathematics and statistics. Now, I don't mean they're doing anything fancy. These guys are not doing like, multiple regression or linear path analyses, and all of the fancy things you learn about in higher level statistics classes. They're doing something very simple. They're paying attention to what predicts what. Now the reason I say this is counter intuitive, is we don't really think of babies as doing this, you know, doing equations. On the other hand, it makes perfects sense, because we know that what our brains are, what brains do, is they track regularities. This is how synapses wire up together, and fire together, is by tracking what events predict what other events. And so given that we know that about brains, it's actually not that crazy to imagine that that's what babies are doing. So, what makes us think babies track statistics? Well, to tell you, I need to tell you about an interesting aspect of language that I'm betting that most of you have never thought about before. But it's something that you deal with every day of your life. And that is, that we don't put silences between words when we talk. So unlike written tests, where there are little nice white spaces between the words, when we speak, there are no little white spaces. And you've all sort of experienced this at some point, because you've been sitting at a restaurant and you've overheard the Hmong family right next to you speaking in Hmong. And you think, "Wow, they talk really fast in Hmong." Once in a while they breathe. Now of course that Hmong family, if they didn't speak English, would be listening to you and thinking, "Wow, they talk really fast in English," and once in a while they breathe. Because we don't actually put pauses in between words. And this is a really interesting phenomenon, because how is it that babies ever figure out what the words of their language are when they're always hearing them all squished together? And we do this when we talk to babies, too. We don't put silences between our words when we talk to babies. So, this is a great learning problem. Because we know it's hard. Any of you who have ever played with computer speech-to-text synthesizers know that it's actually a computationally really difficult problem. But babies solve it by about seven or eight months of age. So, it turns out, and I'll give you one example of this from my lab, but it turns out that if you give infants fluent speech, continuous speech without word boundaries, and you repeat words in it occasionally, by about seven or eight months of age, babies will recognize those words later on when you play them in isolation. So, babies are able to solve this problem. How do they do it? Well, this is where we think statistics might play a really important role. So, I'm going to give you a really simple example from English to tell you why statistics would help you solve the problem. So take a lovely two-word sequence like "pretty baby." Something a cute baby might hear pretty often in their environment. Well, one way that you could figure out that there's a boundary between pretty and baby is to compute some probabilities, and to actually figure out, okay, when I hear "Pre" what tends to come next? Well, I hear "Ty" a lot after "Pre." And it's kind of interesting to stop and think, let's take the syllable "Pre." What other words have "Pre" in it other than "pretty?" Group participation. What other words have the syllable "Pre?" Produce? Different vowel. But it's a good idea. ( inaudible ) Hmm? Predict. Prepare. Precocious. Pre-school, different vowels, close. Hmm? Preliminary. You can get pretty interesting with this. You can say predilection, prevaricate. A lot of these words babies don't hear. But pretend, predict, pretty, these are reasonable things. The point here is that if you hear "Pre," and you were making a bet of what was going to come next, "Ty" would be a pretty good bet, especially if you're a baby under the age of one, because that's mostly what you're going to hear. Now contrast that with after you hear "Ty" how likely is "Ba" going to come next? Well, not very, because the problem is that after "Ty" any sound pretty much can happen. Pretty shoes. Pretty mama. Pretty potato. After "Ty," because you're at a word boundary, pretty much anything can come next. I just did pretty much. "Ty" much as another one. So the point here, and I'll give you a clearer example of this in a moment. The point here is that there are statistical properties of language that fall out of the fact that words, the sounds and words always go together. If you're saying pretty then "Pre" is always going to be followed by "Ty." But what happens at word boundaries is totally accidental and depends on what word comes next. So let me formalize this a little bit for you. Going back to the same example, what one can do is to actually compute the probability that "Pre" is followed by "Ty" by doing something like this, by just asking how often you hear pretty and dividing it by how often you hear "Pre." And if most of the time you hear "Pre" it's inside pretty, that number is going to be pretty high. And the hypothesis here is that this probability will be higher than the probability at a word boundary. And that that could be a useful cue to a learner who is trying to find where words begin and end. Now, lest you think I am making this up, I actually used this example for years before I went and computed these probabilities, because it just seems so obvious that this would work. And I went to a Corpus of Speech to babies under a year of age, about a million utterances, and just computed these probabilities. And the answer is actually interesting and a little embarrassing. So, I'm going to get you guys to help me guess. What do you think is the probability that if you hear "Pre" in speech to babies under a year of age, that "Ty" will come next with the highest possible being 100% and the lowest being zero. How likely is it that "Pre" will be followed by "Ty" any guesses? Eighty? Ninety? You guys are good. It's 80%. Here's a really interesting one. Same thing. What's the probability that if you hear "Ty" that "Ba" will come next in this Corpus? Seven? Twenty? Ten? Higher? Fifty? ( laughter ) As I suspected, it was a particularly well-chosen example. It's a little too good of an example, because in point of fact, you would never usually see quite as an extreme here's a good probability, here's a bad probability for many reasons. But the point is that these sorts of statistics live in the language that babies are hearing. And so what we wanted to do was to ask whether babies can actually make use of that information. So, how would we ask that question? I'm coming back to methodology here. This is work I did in graduate school, but we've done many such studies since. And so I'm going to use this old work sort of as an illustrative example because it's clear. What we did, we wanted to see if babies could do statistics, or track statistical properties of speech. But we didn't want to just throw them into, like just listening to Spanish, or something like that, because there would be so many other types of information that would be in there as well. So what I did was to create a made up language. We call these artificial languages. This is a very small language.

It has four words

tokibu, gopila, gikoba, tipolu. And what I did was I fed these words in random order into a speech synthesizer that did not know where the words began and ended. So, it sounded something like this. ( words all strung together ) Got the words? And you're looking at them. So babies heard this in random order for two minutes. They don't mind this sort of study. You guys would mind it more. ( laughter ) And importantly, the only cue to word boundaries, the only information that could tell babies that tokibu goes together and gopila goes together, were the statistics of the speech, which sounds followed other sounds with regularity. So babies listened to this for two minutes. Then we have to test them. How do we test them? If it was you guys, I could ask you a question. These are eight-month-old babies. I can't ask them a question. So, what I do is a take advantage of a nice, factory-installed feature of babies. And what is that? That is that babies, like you and I, like to listen to familiar things, but only to a limit. If you play them the same thing too much, then they get bored, and they want to hear something new. Again, just like us, this is factory installed. We see this across every species in the book. So what we do is we test babies. I'll show you a video clip in a couple of minutes of how we do this, where we tested babies by playing them, on some test trials, sequences that were words from this little language like gokiba, or whatever it was, tokibu, that's a good example. We played them "tokibu, tokibu, tokibu," and measured how long they stayed interested. On other trials, we played them sequences that spanned word boundaries, something like "bu-giko," which would be the end of tokibu and the beginning of gikoba. They've heard this. But it doesn't have the good statistics of a word. And what we do is we measure how interested they are in listening to these different types of test items. And I'm first going to use my hands and try to walk you through how we test babies doing this. And then I'll show you a video clip. So, just to explain. Here is my lab. Here is me, a younger version of me with less gray hair. Here is a parent bringing their baby into the lab. And the way we test the baby is, again, I showed you a little cartoon of this earlier. The parent is sitting in the middle of this booth. It's a sound attenuated chamber, so it's sounds freakily quiet. The parent wears headphones, airline headphones, that really keep sound out playing music, so the parent can't hear what the baby is hearing and somehow bias them, swinging them to and fro. And then, in the front of the baby, there is a little video camera that is actually hooked up to a TV screen outside the room. And there is an experimenter out there, you'll see this in the video, who monitors what the baby is doing. And the way the study works is that for the first two minutes, the baby just sits and listens to the strain of continuous sounds for two minutes. Towards the end, they start to get a little squirrelly. They get a little bored. Then the room goes silent. And the light in the center of the room starts to flash. We use that light to tell us when the baby is ready, to make sure the baby isn't trying to eat her socks, or something like that. So when the baby is fixating on the center light, then we hit a "go" button. The experimenter outside the booth hits a "go" button. And a light on one side of the room will start to flash. Babies like flashing lights. It's another convenient factory-installed feature of babies. So the baby will turn his or her head to fixate the light. And when the baby looks in that direction, the speaker underneath that light will start to produce one of our test items like "gopila, gopila, gopila," and will continue as long as the baby fixates that side of the room. When the baby looks away, the sound, we give the baby two seconds to look back. If the baby doesn't look back in two seconds, the light is extinguished, the sound is turned off. And the center light starts to go again in the silent room. So, what is happening here is the baby is the boss. The baby controls how long we play each test item. And in that fashion, we can get a sense for what the baby is most interested in listening to. So what I'm going to do now is show you a video clip of an actual baby in one of these experiments. I apologize. It looks a little drunken at the beginning. It's an old video camera, and it takes us a while to zoom in on the baby's face. But I'll do a voiceover for you. ( words strung together ) There is our experimenter, who can't usually hear what the baby hears. Okay, now the center light is on. The side light. The baby is listening, interested. The baby looks away, so we tell the computer "baby looked away." The center light is on. Side light is on. ( string of words continue ) The baby looks away, but he comes back, so the computer keeps going. ( words stop ) And he's done. So the babies actually do a pretty good job for us, of showing us what they're interested in listening to. So, just to put a fine point on this, these are the stimuli. This is what the baby is faced with, okay? This is a stream of speech. And here is what we're hypothesizing that they learn. That they actually figure out which sounds cohere together into something like words, not that we really think that eight-month-olds really know what a word is. And then we test them on something like tokibu, which is a word from this speech with really good statistics. And we contrast that with something like bugopi, something, again, that they've heard, but it doesn't have as good statistics, because "bu" could have been followed by a lot of other things in this speech. And at some level, I was really surprised by these findings. At another level, I figured it had to be this way. Babies can solve this task. So, we and many other labs have demonstrated that when infants are confronted with fluent streams of speech like this, they readily track their statistical properties, as well as many other features of speech. And it turns out that this idea about tracking statistics is not limited to something just like finding the words in speech, which even though it's obviously a really important problem. If you can't find the words then you're up a creek. You can't learn anything else without your language. It turns out that these statistics are really important for many other aspects of language learning, which I won't take time to go through now. But which we and others have been actively exploring. It also turns out that this isn't just for language. So these same learning processes operate across a lot of other domains of experience. So, one line of work that we've done looks at how babies track patterns of statistics in music. We played them something that sounds like this. ( multi-tonal melody ) Okay you get the idea. Stop! And they can track the statistics of pitches just fine. They can track the statistics of objects. They can track the statistics of streams of action. And at this point it seems almost limitless, the kinds of places where tracking what's predictable or reliable could turn out to have important ramifications. So, this is something that we're actively excited about. Now since I started by telling you about work that was published in 1996, I thought I would turn next to tell you about a study we're doing right now, that turns from asking whether babies can track statistics. In the ensuing decade, I think we've all been quite convinced that babies can track statistics. And now we're trying to ask, well what are they for? Okay that's fine that they can do this. But how do they use them? What do they do with them? And to tell you about this work, this is work being done by a graduate student in my lab, Alexa Rhomberg. And what she has become interested in is whether babies use statistics to make guesses about what comes next. So one thing we know about adult language processors is that you and I are constantly making guesses about what is going to come next, in real time. So, as you guys listen to me talk, your brains are actually taking bets and making guesses as to what sounds, words, types of words I'm going to say next. And there is very nice evidence that this is going on in adulthood. And this is kind of amazing. But it also helps to explain why it is that I can talk as fast as I'm talking, and you guys can presumably, hopefully, follow it and make sense of it. As I talk, you're laying out your cards, you're making guesses as to what words, sounds, types of words are going to come next. So what Alexa wanted to ask was whether infants are also making guesses in real time. Are they expecting what is going to come next? So here is what she did. She took 16-month-olds, which for me are really old. That's like old babies. I'm used to eight-month-olds. And she gave them exposure to very simple pseudo-English sentences. And here is the manipulation. I'm actually going to lay this whole thing out to explain it. On some of these sentences, the adjective predicts the noun perfectly. We call these deterministic trials, because you know exactly what's going one come next. Every time the babies hear "little," "dog" is going to come next. And every time they hear "pretty," "fish" is going to come next. So they could, in principle, make a bet, when I hear little, dog will come next, and expect it. On the other hand in these trials, "happy" doesn't tell you whether "kitty" or "birdie" is going to come next. Either one could occur with equal frequency. And so here, hearing the adjective actually is totally uninformative for you. Here's what this experiment looks like. Oh, yes, so here is the question. Will hearing the adjective for these deterministic cases help the babies predict what noun will come next? Can they learn enough about these sequences to make bets as to what will happen next? So here are the materials from this study. Babies are first familiarized with something that looks like this. If you kind of watch carefully, you'll see one of the pictures flash as you listen. >>

Computer

This pretty fish. One little dog. This happy birdie. One happy kitty. >> Okay, so what we're doing here, for two minutes is teaching the babies both about the adjectives predicting the nouns, so which adjectives go with which nouns, and also where on the screen the different nouns will appear. Because what we're going to do next is see whether they look to the right place when they hear the adjective. So that's the familiarization. Here's the test. So we start, just to get their attention in the middle of the screen, with a little snippet from "Baby Einstein," because it gets their attention. Notice the screen is blank here, until the noun. >> One happy kitty. This pretty fish. One little dog. >> So what we want to do is to track their eyes, to see whether their eyes look to the place on the screen where the picture of the noun will occur before it happens. So right now, I'm going to show you a video of an infant in this task. And I want you just to watch that infant's eyes as I play this video. So here is a real live baby in this task. Well, not live right now, but previously live. You can't hear the speech, but watch his eyes. ( no audio ) So what we do is after these experiments, we record the infants in this fashion. And then we go frame by frame, every 30 milliseconds, or my undergraduate assistants do this, and keep track of where the baby is looking at every given moment. And here's what we find. Now before I show you the data, I'm just going to tell you what you're looking at here. Right here, this is zero milliseconds. This is when the adjective is spoken, so the baby hears "little" right there. And they don't hear the noun "doggie" until way out here. And our question is what do babies do? Do they know where to look when the screen is blank. So, the screen is blank all the way here, until the noun goes on. So, chance is 50% just for argument's sake. So first I'm going to show you the babies in the probabilistic condition. These babies have no way to know where to look because "little" is predicting both the right and the left side of the screen. And what you see, oh, I'm sorry. What we're most interested in at this period of time, when the babies have heard the adjective but haven't heard the noun yet, and haven't had a chance to react to it. What you see is that for the babies where the statistics don't tell them where to look, they're right on chance. They're looking half the time to one side, half the time to the other side. And only when they hear the noun and that noun picture flashes on the screen do they know where to look. The interesting condition is what happens when the statistics tell you where to look. And it turns out that the babies can use them. They're not using them wildly. They're not up at 100%, but they are statistically reliably making a bet as to where to look, and making the right bet. They're looking into the right place long before that noun is spoken and long before that noun is pictured. So this suggests that babies, like you and I, are making bets in real time, as speech is going. And we have a very active research program right now to try to explore the nature of these sorts of predictions that babies are making. Now the last thing I want to talk about actually isn't about my own research. But it's about sort of some of the educational implications of what people in my field have been studying and trying to find out about how these great babies and children learn, as opposed to older learners. And so, many of us are interested in second language learning, because it's ubiquitous and it's a struggle for many people. I don't know about you guys, but I had a terrible time in high school and in college trying to learn foreign languages. And many people do. So why is that? So what we want to know is when is the best time to learn a foreign language. How should you teach it? Are there detrimental effects in early childhood and infancy of being exposed to multiple languages? Are there positive effects? Well, the clearest data in my mind have to do with the timing. There's clear data on three of those questions. And here is a graph from 1990. These data have been in the public domain now for almost 20 years. And I'll tell you why I think it's shocking that they've been around for so long, given how our educational system is set up. This is a task. This study was done with individuals at the University of Illinois who had emigrated from Asia. And they were college, grad students, or faculty. So these were all people who were using English in their daily lives. They're engaged in an English speaking community. And it was a test where people were given sentences. They were either grammatically correct or had a grammatical error of some kind or another in them. And they had to identify if there was an error, if the sentence was good or not. All of these subjects had been using English for at least 20 years at the time of testing. And what you can see, what's really interesting about these data, these are native speakers, so these are people who were actually born in the United States speaking English. So this is about as good as you can do on this test. And you can see that individuals who immigrated to the United States between the ages of three and seven, they look native. They're doing just fine. But amazingly, you see a huge drop off, people who arrived at 8-10, 11-15, or 17-39. And here is where the educational implications come in. When do we start teaching foreign languages in our schools? Maybe here if you're lucky. I have a kid who is in kindergarten in the Madison Public Schools, and it just makes me nuts that there is no foreign language instruction in the elementary schools in Madison. So, this is actually, this is really actually an appalling thing. And I think it's a real public policy disconnect between what research has shown. Again, this paper was published in 1990. A real disconnect between what research shows and what our educational establishment has invoked. So, earlier is good. Earlier is better. So what do we know? The earlier the better, probably. How? Immersion seems to be the best way. And we can talk more about that in the question period. I want to try to wrap up. Interestingly, people often worry that there are detrimental effects of having babies be bilingual or exposing infants to multiple languages. It turns out that their aren't. There are brief delays. Some infants who are in a bilingual environment will tend to lag monolingual infants by a couple of months in all their linguistic milestones. But if you think about it, they're only getting half as much input of each language, right? Like my kid hears English all the time, and your kid hears English half the time and Spanish half the time, the fact that their only behind by a month or two is actually pretty amazing, right? Because they're only getting half as much English, and yet they're doing fine. And the kids catch up just fine, no problem. And in fact there is some evidence to suggest that there are real positive effects of bilingualism. Namely, a level of cognitive flexibility. If you speak two languages fluently in your daily life, you have to be capable of going back and forth between them effortlessly. And it appears that children who are bilingual, actually probably the brain structures that subserve that ability to flip back and forth, seem to help them with other types of tasks that require, you know, changes of your mental state or changes of set, allowing you to be more cognitively flexible. So one thing you might be wondering, you know, this is a science talk, and I'm in psychology. And people usually like to know, well what about the brain? What is the brain doing? Can you do those fMRI things that are so pretty? No, we can't. We know almost nothing about the developing brain as it pertains to language. We can't do fMRI because babies move. So if you go in an fMRI magnet, you have to be still for like an hour. I don't know any six-month-olds who can do that. And much of what we know about the developing brain, other domains, we know from non-human animal studies. But of course non-human animals don't develop the same types of languages that we do. So we're actually pretty hamstrung. This is an interesting figure for many reasons. And I won't really go into what it's about, except it's about the distribution of gray matter versus white matter. It doesn't matter. ( light laughter ) The reason I put it up here is look where it starts. It starts at age five. So, a place that I think will be very interesting, maybe in the next couple of decades, is there are technologies that are being developed that are non-invasive, that don't require you to be still for an hour, and that might serve a really useful role in trying to begin to understand the brain underpinnings of all of this. But for now, I think we know remarkably little. And again, with respect to public policy, I think this is interesting because you hear all this stuff about you know birth-to-three, and everything we know-- all the important stuff for the brain is going on then. But I don't think that can be true. And here is a really nice quote from John Bruer, who is the Director of the McDonald Foundation, which funds a lot of neuroscience research. And here is what he says. "Apart from eliminating gross neglect, neuroscience cannot currently tell us much about whether we can, let alone how to, influence brain development during the early stage of exuberant synaptic formation." That's when we have all these excess synapses, when we're very small. "The brain is not 'cooked' by age three or age 10. Our brains are remarkably plastic and we retain the ability to learn throughout our lives." I take that as a very hopeful type of message. But it also suggests some caution. Any of you who are in the early childhood field, or the elementary education field, you hear a lot these days about brain based curricula. People are always trying to use the brain to justify things. You see it in the paper all the time. We don't know hardly anything about the brains of these developing critters. And so, I would just urge you to exercise caution. So, to conclude, you already saw my conclusions. Infants are extraordinary learners. They don't need teaching in the sort of old-fashioned sense, or over stimulation, going beyond what their worlds provide naturally. Much of what they learn, we couldn't teach even if we tried. I think the word segmentation example, that finding words influence speech is a great example of that. Those of you who have kids, your kids all learned to do that, you probably never though about it as a problem. They figure it out by themselves. And I think that's a really important message about children. But at the same time, they really, really need experience. It's not that they can figure it out themselves if they have nothing to learn from. The absence of being exposed to a language, be it a spoken language or a sign language, they can't do it by themselves. They can't just make up languages. They need that experience. And they need an interactive experience to learn. It turns out that they don't learn as well from videotape as they do from a live interaction. And that makes sense for many reasons. It's true of many other species as well. So I think our job as scientists is to continue to try to figure out the nature of these learning abilities so eventually we can help kids where it's not going quite right. So in my own lab, we've started research projects looking at a number of different populations of children who are suffering challenges in learning language, including children who are deaf and wear Cochlear implants to try to hear, and also children who have diagnosed language disorders. By studying how they learn, we can maybe eventually be able to help them. And finally, I think our job as educators and caregivers is to provide the sorts of experiences that are going to be good for learning from. I don't think we need to bombard kids with a lot of extra stuff. I think just by getting down on the ground and talking to them, we can get really very far. So let me end by just thanking the people who fund the research that we do, and the people who actually do the research that we do, the members of my lab without whom we couldn't do anything. And then most of all, I have to thank the little people who make all of this possible. So thank you very much. ( applause ) Yes? >> Has anyone replicated the word separation study with other species? >> That's been replicated with cotton-top tamarins. And it's actually been replicated with rats. And so rats can do the word study as well. Yes? >> How about adults? >> Yes, adults can do it. We use different methods with adults. We don't have them turn their head. We give them a tool alternative choice test, where we play them things and then say "which of these sound more like what you heard before?" And they can do that just fine. We don't have a good way of saying whether they're better or worse at it than the babies, because the babies' test is sort of easier. They just have to turn their head and listen to something. There is no memory component there. But adults can do that test just fine. Yeah? >> Is there evidence that shows that babies learn language, more than just classical conditioning? >> Well, I would say that all of this suggests that it's not classical conditioning. Because in classical conditioning, you have a reinforcer, so there is some sort of reward that is used to link the stimulus to the reward. Here, there's no reward. I mean, the only possible reward I could imagine in these studies is that somehow the inherent reward of learning. There might be some dopamine pathways that get all activated because you've learned somehow. But there is no one out there who's like dangling the meat and ringing the bell. It's really being driven internal to the infant. Does that make sense? Yes? >> Do you or any of the people in your area look at the work of Montessori, the work she did in the early 1900s, especially around the concept of sensitive periods? >> In fact, these sorts of data, yes, these sorts of data--. Go back. This curve I showed you, that's a classic sensitive period kind of curve, right? Because what you see is that there's a time window during which experience will have a larger effect. You can give the same experience out here and it's going to have less of an effect. Yes, absolutely. And there is a lot of work on sensitive periods in areas like birdsong learning, and biological domains, where it is easier to manipulate your subjects than it is to manipulate humans. Yes, absolutely. >> While you've got that chart up, is there any evidence about language learning from ways that aren't immersion, because obviously that's the age that immersion started, so is there any other evidence about the effectiveness of other methods? >> You know there is an entire field of applied linguistics that try to fight about which pedagogical techniques are the best. And so I'm not in a position to say. I mean, I know in terms of things like school based language programs, the ones that seem to work quite well in Canada, where this is just ubiquitous, are dual immersion language programs, where you spend part of the day in one language, you spend the other part of the day in the other language. And there are native speakers of both languages in that environment. There is a charter school in Madison using Spanish in that fashion. And it seems to be working in that environment as well. Beyond that, you know, I don't really know. Yeah? >> Could you just tell us at what age that starts? In Canada and the one in Madison? >> In Canada, as early childcare. We spent a year in sabbatical in Montreal when my son was two. And the daycare was dual immersion bilingual. And by the end of the year he was fluent in French without trying to be. In Madison, the Nuestro Mundo Charter School, it's a public charter school near Monona. They start in Kindergarten. >> How are they doing starting at age five? >> Well, based on these data, you know, you would expect that they would be fine. Of course, it probably depends from child to child on just how much exposure they're getting. But based on those data, I would anticipate that kindergarten would be just fine as a point of entry into a language system. Yeah? ( inaudible ) Can you repeat the question, please? >> There's a lot of talk about phonics and awareness. Is that something that's related to what you study? >> Yeah, we kind of looked at that in our own studies. I'm sorry, the question is, there is a lot of talk these days about phonemic awareness, which for those of you who don't know is the ability to sort of piece apart sounds prior to learning to read. So being able to say here is the word cat, now say cat without the "C," and the ability to do things like that requires some knowledge about how the building blocks of language work and what the sounds are. We haven't looked at that in our own research, because we're looking at babies. And so we don't really have any independent way to look at their knowledge of their phonemic awareness beyond precisely these types of studies. But I'm sure that there are individuals out there who are engaged in these sorts of learning studies, who are manipulating those sorts of factors. Chuck, do you know if there is anyone doing anything like that? >> No, I don't. >> There's a reading expert right in front of you. So I would feel uncomfortable speculating in Chuck's presence. Yes? >> Your child that learned French fluently, how long did it take him to forget it? >> Well, that's a really interesting question. If you ask him now to speak French, he claims he doesn't speak French. He knows the alphabet and things like that, and counting words. But there has been some really lovely studies on this issue I talked about earlier in the talk about how we lose our ability to perceive certain sounds with experience. And people have done very interesting studies looking at children who were in one language environment as babies, and then were adopted and put in a totally different language environment. And then they're tested, say in their 20s, and it turns out that for something like that Hindi "dah" and "dthah" thing that you and I can't tell the difference between, if you heard Hindi when you were a baby and you heard it a lot, then even though you haven't spoken it for 20 years, you can distinguish between those sounds better than you or I can. So it's forgotten, but it's not lost if that distinction makes any sense. >> The reason I asked the question was that my brothers were born in Mozambique, and they spoke Tsitwa. And in the three to four months that it took for the family to come home from Africa, they completely lost their language. And that's why I asked the question. >> Yeah, I think that's fairly typical. Unless you have some supports to maintain a language. I mean, the pathetic thing is, you know, my son had this great French, but we were unable to sort of get it together to find French speaking babysitters, and so we didn't maintain it. And if you don't have peers or other caregivers speaking the language, you won't remember it. But I'm willing to put money on the idea that if he takes French in high school, because he won't get it until then, maybe middle school, that his ability to re-acquire French will be faster than my daughter who really was too little and didn't really learn any French when we were in Montreal. Yeah? >> There is a belief that if you learn two languages, one goes on one side and one on the other, especially when you have a stroke that you may lose the one and gain back the other. Does that have an effect, with your babies of looking which way depending on which side of the brain it's being recorded on? >> Should I repeat the question? So, there is some evidence to suggest that if you learn two languages, they'll go to different sides of the brain. And he's asking whether there is any effects on which side the sounds are playing in our testing rooms. I think the evidence suggests actually that when you learn a second language it doesn't go to a different side of the brain, but that there might be sparing because say any given stroke is only going to take up say so much of your left hemisphere. And so the more language, if you speak more languages then probably more of your cortex will be colonized by language. But I don't think there is any evidence to suggest that say your first language is on your left and your second language is on the right. It seems to be that depending on how early you learn the language, the cortical areas will either overlap each other or will be very near to each other. With respect to the second part of your question, we always control our studies so that half of the stimuli are on one side and half on the other, and it's all, you know, balanced out in all of the scientifically rigorous ways. So if there were any effects about side, they would wash out of the data so we wouldn't be able to see them. We certainly never observed any. Did you have your hand up? Yeah? >> I have a weird suggestion. We know that different languages actually have different melodies. Isn't that true? >> That is absolutely true. >> Now my thought is that as an adult, what difficulty we have in learning a new language is that we're stuck with our own melody. >> That's true. >> And if that is true, insofar as that is true, it occurred to me to do the following experiment. Get a group of adults who are willing to learn a second language, let's say French. But, the way they will learn it is by singing it. You're not going to do it through ordinary speech. So Frre Jacques would be one example. And whatever you learn will be through melody, but not the American language melody, because you're singing different songs. And I'm just wondering what would happen in the ease of acquisition? >> Should I repeat the question? So, here's the hypothesis and here is the experiment. The hypothesis that is the part of our challenge in learning a foreign language is that the melodies of languages vary from one to the next, which is true. And the experimental suggestion is to ask whether people would learn language better if we sort of did away with the different melodies of the language, and taught the language by singing. So you would essentially use melodies that are not tied to a particular language. So, learning French by singing Frre Jacques. I'm sure people would learn. The problem is that they wouldn't learn the language, because such an integral part of the language is its rhythm, and its pitch, and also its meaning. It would be very hard to learn meaning by singing. But that said, we actually have done some experiments looking at how well babies learn from sung speech versus spoken speech, and how well they learn from speech that has lots of pitch versus just a little bit of pitch. They learn much better when there is a melody. Much, much better. So while I don't think your strategy would help you learn a full language, because by singing you distort so much. When we sing we really distort the sounds of a language, even within English. If I sing something, I'm really distorting the sounds of English. I don't think you'd learn the language, but I think you'd learn a lot about the words and the patterns that you were trying to train them on. So, yes and no would be my guess. Yes? >> You were talking about "Ra" and "La." Is this related to the way your tongue moves against your palate. My mother could never say "L" or "La." My father could never say "Da." Are these learned things in terms of your tongue when you hear it? >> The question was whether the way we learn these sounds is tied to how we produce them. I think the answer there is yes and no. In the data about learning that your language doesn't use certain sounds by the time you're six months or a year. Infants aren't usually producing those sounds yet. So some of this learning is probably independent of production. With that said, when people do retraining studies, there is a lot of research right now doing things like trying to take Japanese speakers, and how do we teach them "Ra" versus "La." And there, training on the motor part does seem to improve the way you hear things. So, I think there may be some really different mechanisms going on for that original learning that happens in infancy, which probably is not tied to production. Babies, our productions in infancy sound pretty much the same the world over. For the first six or eight months or so, we babble the same sounds. And then they only get tuned up much later on. So it probably doesn't help there, but it may help when you're trying to relearn a sound that you've lost, or when you're trying to rediscover an aspect of language that you no longer use. A question in the back? >> Do babies learn tonal languages more rapidly than say Western languages? >> That's a great question. Do babies learn tonal languages more rapidly than Western languages. For those of you that don't know, a very sizable number of the languages of the world use pitch in a way very different than English. So, a tonal language is a language where the pitch, and the pitch contour of a word totally changes its meaning, not just shades of meaning. Do we have any speakers of a tonal language here? Mandarin or Cantonese or Thai? Yeah? Would you give us an example? Which language do you speak? ( inaudible ) Can you give the "ma" example? The sound "ma," with all the tones. ( inaudible ) >> There's mother, "ma." And there's "mah," which is a kind of fabric. And there's mah, which is coarse. And there's "mah," which means you're cursing. ( laughter ) >> Thank you. You all heard that? So this is a really interesting challenge for an infant learning language. Sorry, I have a cold, I'm losing my voice. If you're learning English, you have to learn to ignore the stuff that he just said. Because "cat" and "ca-at" and "cah-at" all mean the same thing. But if you're learning Mandarin, you have to learn to pay attention to those things. In answer to your original question, we don't know which of those languages is acquired easier or harder, because at some level, all languages have hard things and all languages have easy things. But we're actually just about to start a project in my lab to see if we can teach English learning babies to pay attention to tone and see if we can teach Mandarin babies to stop paying attention to tone by manipulating the statistics of the input. >> Is there anything you can do prenatally to affect language learning? >> Well, to affect. We know that in the uterine environment fetus' actually do learn something about the rhythm of their mother's language. So when babies are born, they will listen longer to speech that has the same rhythm as what they heard from their mother in the uterine environment, as opposed to sounds that have different rhythms from what they heard in the uterine environment. So, it does suggest that they are learning from what their mother says. With that said, I mean, the way to think about the uterine environment, I don't know if any of you guys ever watched the "Charlie Brown" or the "Snoopy" movies on TV long ago. And the parents and the teachers sounds like, "Mwaa-waw, mwaa, waw, wa..." That's what it sounds like in there. ( laughter ) And so even if you wanted, there are all these products now, like speakers that you can plop on your belly. I mean you could do that, but I don't think they're going to get anything other than rhythm. And rhythm is great. But I don't know why you'd spend hundreds of dollars to buy speakers to teach your babies Mwaa-waw, mwaa, waw, wa. ( laughter ) So, I don't think we need to start doing a lot of inter-uterine training. In the back. >> I live in co-housing. And Sunday evening I sat at the table with two tenants who spoke Mandarin. The woman said "ma" can mean

five different things

"ma, ma, ma, ma, and ma." I said that's easy and I copied her. She said that's perfect. I know nothing in Chinese. It was interesting you chose that example. >> The question is whether you would remember it. >> I don't know which of five. >> Right, exactly. ( laughter ) Exactly. And it turns out that speakers of tone languages actually represent pitch somewhat differently than we do. So here is a cute experiment that colleagues at UC-San Diego did. They brought in speakers of tone languages and languages like English and they had them say some words. And then they had them come back a week later and say the same words. And the speakers of Mandarin said the words at exactly the same pitch a week later, exactly, within a semi-tone, exactly the same pitch. The English speakers when they came back and said "cup" and "cat" again, totally different pitches. So, there really is, these tone languages are shaping up tone language speakers' brains to process pitch somewhat differently than those of us who speak languages like English. Yeah? >> I'm an early childhood educator, so when I read research or go to talks I always look for the practical application, to bring it back to the classrooms that I work with. And I think that there is a lot of practical application for your passive learning work, particularly because I think there is issues with quality of care in environment, especially for infants. And so I was just curious how you as a lab connect your research to sort of the greater profession out there. Do you do anything with some of the programs in Madison? I'm new to the area too. >> Okay. We actually don't currently do anything with any of the programs. Oh, I'm sorry. She said she's new to Madison and she works in the area of early childhood education, and she's always looking for the practical applications of the work that she reads about or hears about. And she find these ideas of passive learning really important, potentially, with respect to things like quality of care and is my lab collaborating with any sites to try to implement these? We are not, but I would be very open to talking about it. Because my lab is at the Waisman Center, and there is an early childhood program at Waisman that both my children have attended, I have always wished that I could be more involved. But I never made quite the right connections to do that. I agree that quality of care. I get very worried when I hear about childcare settings where the babies are just propped in front of a TV and, you know, that's exposure and experience, but that doesn't strike me as any sort of surrogate for what kinds of natural experiences infants should be having. So I'd be delighted to talk more to you if you have ideas, that would be great. Yeah? >> If you want a child to learn something outside of let's say, French or Spanish, that you maybe learn in America. I mean, where would you talk them. Are there any schools that basically, you know, Hebrew, Russian, languages that they don't speak. >> There are. For example, there are. Yes, I'm sorry. The question was if you want your child to learn a language that's not the usual French, or Spanish, where would you take them. In Madison, there are several Chinese schools, actually. And I've heard they're very good. I've never explored them myself. Hebrew is something that actually, my son goes to a Hebrew school club that's not religious. He's learning Hebrew as part of his cultural, to develop his cultural background. So there are things like that. The great thing about Madison is that if there's any language you're really interested in, there are 18 graduate students studying it who are underpaid, and they would be happy to come play with your kid. This is actually a great community, I think. Just go to Eagle Heights and hang a flier, and you'll have all the language speakers that you want. But I don't know of any formal environments where some of the other languages one might be interested in. Yeah? >> Would you comment, please, about this movement to teach infants, babies, American Sign Language. >> The question is would I comment about this movement to teach babies American Sign Language. To be clear, they're not teaching them American Sign Language. They're teaching them signs. It's called Baby Signs. You might have heard about it. Basically, what it does is you teach your baby a small subset, maybe 20-100 individual signs, not in the context of sentences. So more, milk, eat. Stuff that might allow your infant to be able to communicate with you a little earlier and make them less frustrated. So the reason that I distinguish that from learning American Sign Language is ASL is actually a rich, full language, as rich as English, as rich as any other language. These programs don't do that. They're not training the kids, not giving them experience with full, rich ASL, which is actually extremely difficult to learn as an adult, having tried, but very easy for children to learn. You know, I taught my kids a few signs, because I knew them. And I have to admit, my kids' first word, both of them. ( laughter ) So I do think it allayed a little frustration. Now what I don't think is right is some of the people who market this stuff, claim that it's going to give your kid an earlier onset into learning English, or learning language. I don't know that that's true. I don't see any evidence for that. But I don't think there's any harm in giving them the tools. And actually, to be honest, if I were king of the world, I would legislate that every child had to learn American Sign Language, because that way, some day when you lose your hearing, or you have a child who's deaf, or something like that, you don't have to have this horrible thing of, I can't sign. I actually think we all should be learning ASL. Not as adults, because I can tell you having tried it as an adult, it's really hard to learn mirror image. When you learning ASL, you watch people do stuff, and you have to mirror image it. But when you're like two, no problem. They're fine. So I think that's a really interesting question. Yeah? >> I didn't hear your answer, whether it's good or bad for small children to be in front of television. Does it hurt anything? >> The question is, is it good or bad for small children to be in front of a television. Is it good for them or is it detrimental. I mean, I would say that anything in moderation is fine. You can't go through the world and tell people to turn off their TVs. That said, the mere fact that there's now a satellite TV channel 24/7 for babies, I find profoundly disturbing. I mean there's a satellite TV channel just for babies, 24/7. That scares me! So there is a little bit of evidence that it can be detrimental in the sense that time spent watching TV is time not spent on the floor, eating dirt, and you know, taking the pots and pans and banging them around. And there's not much evidence for learning language, learning much rich language from anything like TV. It's pretty hard to learn when it's 2-D. Kids have trouble understanding two dimensions. They don't always understand it when they see something that's two dimensional, that it represents something that's three dimensional. That's very difficult for children. So you know, do I let my kids watch TV occasionally? Yes. Would I subscribe to a 24/7 cable channel TV channel for babies? No, and I find that really upsetting that there's a market out there for something like that. >> People who stutter don't stutter if they sing. >> Right. >> Now that suggests, again, that music or melody has a very important connection to language in some way. So I'm pursuing my weird suggestion. ( laughter ) I would think it would be very interesting for people interested in language development to experiment with music. I liked your answer, yes and no. Because we know so little. But to try to discover how they could just hum the melody of a language, or for example, without words. And we know there are different accents. I mean, where you accent the words. And to then do something where you accent it, and of course combine it with anything else, ordinary grammar training and word training, and so on. But try to use music. >> The question, in brief, was that you're pursuing the idea that music matters, and that you should be thinking about the sort of accenting, and that different languages do this in different ways, and that music might be a mechanism. I will point you to the second slide of my talk. And in fact, we do do many of those studies. We've done a number of studies looking at accent and rhythm, and the fact that infants seem to be really honing in on that. So I completely agree with you. One of my colleagues has a paper where the title is something like When the Melody is the Message. And I really think that in early infancy, melody and rhythm just override everything else. And in point of fact, the challenge for babies is to get past just melody and rhythm to learn all these intricacies about the individual sounds of the language and how they combine. But you know, maybe there would be a good intervention to try with adults. I don't know. We could talk more. I'm on board. Yeah? >> With regard to the curve that you just showed, do you know what happens to it over the next few decades? ( laughter ) >> Yes, I do. I know a little bit, actually. You notice that this is clumped between ages 17 and 39. It flattens out. But here's the really interesting thing, and I didn't point to this. This graph isn't graphed very well, because what's more interesting, if you look at it or do it as a scatter plot, what you see is that at the beginning, when you look at the native speakers and the early learners, if each dot was a subject, the dots are really clumped right around here. But as you get further out, the variance starts to get huge. And out here, you have here you have people who sound like they're native, act like they're native, and you have people who are unintelligible. And so one of the real mysteries here, and something that I don't think anyone has a clue about what the reason is, is why, it's almost like there's two separate learning processes going on. What happens when you learn early in life is totally different than what happens out here, where it's almost more like some people are good at music and some people aren't. Some people are good at art and some people aren't. That's what the data was like here. It looked more like a talent or a skill, whereas up here it looks more like something that just sort of, I don't want to say it comes factory installed, because that makes it sound like I'm saying it's innate, and language can't be totally innate, because we know so much learning goes on. But there's something very primal or very early going on. Yes? >> In regard to television being not so beneficial to young children, I was wondering at what age groups it becomes beneficial for, like for second language maintenance or for vocabulary building. >> The question is whether older kids can get benefit from TV, DVDs and things like that for vocabulary maintenance and language building in foreign languages. I mean, some of those systems, like the Rosetta Stone, I mean, I think there are some products. I actually was playing with the Rosetta Stone recently. I was thinking about buying it for my kids, because I want them to learn Italian because we're going to spend a year in Italy in a few years. I thought about doing that as a first step, because it is engaging. It's kind of cool. And I do think that products like that actually help. And I think once you know a language, it can provide a maintenance kid of thing. When we were in Montreal and my son had just learned French, we went out and got the French video of "A Bug's Life," which he had seen in English before. And he was really nervous, this was his first French video. He was kind of upset about it. And we put it on. He watches for a few minutes. Oh, my God, it's the same thing! ( laughter ) He had this sort of shocking linguistic experience. That just because it was in French, it was the same thing. And I absolutely think, that for kids, once they are already engaged in knowing the language, any materials in that language are going to be, any sort of experience with maintaining those sounds is going to be helpful. Yes? >> There are tribes in Africa that make clicking sounds as a major part of their language. Have you done any studies with those? It would probably not be that easy to do. >> The question is that in Africa, there are some languages where clicks are an important component of the sound. What's that movie, "The God's Must Be Crazy," where there's clicking? I haven't done work on them. But another lab has. And what they show is that infants can discriminate between the clicks, just like they can sounds of other languages. But interestingly, adults can, too. So unlike la versus la, or dah versus dthah, you don't lose the ability to distinguish between those clicks, which is kind of interesting. We think that it might be because the clicks are non-random. They're using up the same bits of the acoustic space as sounds in our own languages, like ba versus pa, but in click sound. And so, we can perceive them because they are, in some ways, extremely similar to sounds in our native language, even though they sound so different. Yes? >> What do you see as happening in children who become autistic and lose their language? >> The question is what do I see as happening in children who are autistic and lose their language at 1-1/2 or 2 years? I'm not an expert in autism. It's a pretty fraught area, so I'm not going to fully answer your question. But I do think it's actually much more common for children with autism, instead of losing their language, they actually have difficulty developing in the first place. So I think it's much rarer. People, you know, you hear of cases where children lose language, but I think it's much more common for children with autism to actually be brought into the pediatrician's office and undergo a neuro-psychological evaluation precisely because they have not been acquiring language at the age appropriate pace. Yes? >> Since it's clear that social learning is so much better than mechanical learning, I should think that there's a nice study developed that would debunk this satellite TV channel. ( laughter ) >> The problem is you can debunk it all you want, but if there's a market, people are going to buy it. I mean, look at this. This data, I'm telling you, this is very, very well-known data. And yet, it hasn't changed a thing in our country about curricula development, to my knowledge. So, you know, there has to be, people have to be pretty motivated to change. And unfortunately, I think in the case of the 24-hour, people have their reasons for doing what they need to do. I think unfortunately in our current society, there are just many cases where parents lack the social or family resources to have the kinds of quality interaction time with their child that the child needs. And TV is one surrogate. I mean, we've all had the experience of needing to take a shower, so putting the kid in front of something. But I can only imagine, I just think for myself, if I were living in poverty and didn't have resources, and wasn't in a safe place, I would use TV a lot more than I do now, where I can send my kids outside to play and they're safe. So I feel a little bad, sort of casting, being tongue in cheek, but I think there are reasons why people do the things they do. And perhaps the thing we should be doing as a culture is trying to address those root causes, and giving kids safe places to be, which would be nice. Yeah? >> There is some evidence that autistic kids can learn a lot better with mechanical language. Have you studied that? >> The question is that there seems to be evidence that children with autism are better with learning from mechanical language, and have I done research in that area. I have not. But there is evidence that children with autism, some children with autism do learn things with mechanical sounds better than typically developing children do. And why that is, I don't know. I don't think it's solely because of something about the social contacts versus the lack of social contacts. There's some evidence that, for example, children with autism have better pitch abilities than children who don't have autism. And so, their auditory systems might be working slightly differently. I think there could be a lot of reasons why we would see those sorts of effects. Yeah? >> Do you know if it's true that children and adults who have musical abilities learn language more easily than those who don't, because they hear the sounds better? >> It goes back to your hypothesis. The question is whether children or adults who have musical ability learn language better. There is no evidence for that. Oh, you are so kind, thank you. Ricola! Sorry, I'm a singer. I don't know of any good evidence. I'm going to talk like this, because I actually can. Down here, I lose my voice. But if I talk up here, I'll sound kind of ditzy, but. ( laughter ) You'll be able to hear me. There isn't actually any good evidence to suggest that musical training helps with language learning. >> I didn't say training, I said ability. >> Ability is hard to measure. The problem is there are measures of ability, but they don't have them for infancy. And it's looking at, the best proxy we have is looking at who becomes, who gets musical training or not, which is only partially related to musical ability. There are probably many, many individuals out there who have musical ability that never have the chance to foster it through musical training. That said, you know, there are some hints out there. I just read a paper today, actually looking at adults with a little language learning tasks that involve some tone language stuff. And it turned out A, that musicians were better than non-musicians at learning this language system, and B, that across both groups, individuals who had larger chunks of cortex in certain brain areas that we know are implicated in sound, learned the language better than those who didn't. And the musicians tended to be the ones who had those better chunks of cortex dedicated to sound. So I think your hypothesis is probably, there's some truth in it. The reason I sort of giggle about it is, I'm a musician and I'm miserable at learning second languages. So you know, my n of one right here always makes me, I should be studying that hypothesis, but then I look at myself and think, then, what's wrong with me. It's a really interesting idea. Thank you. ( applause )