Sound is very important for language development and, interestingly, we have a pretty amazing capability to break up sound in our native language. Now you might think when you're listening to me that what you hear are words, right, and you would be correct. I'm not going to tell you that you don't hear words, you do. I hear words when people speak, but when we actually look at what's produced in a spectrogram, what you find is that there are no clear word boundaries. Now, I learned this because one time we were doing an experiment, this experiment we wanted to create some sentences in which we actually took out a word. And instead of presenting the word that people would hear, we would present a visual word on the computer screen. So I asked a research assistant, to go off and take out these words, gave them a list of all the sentences with all the words that I wanted taken out of those sentences, and he came back two minutes later. And he said, I can't do it. I said, what do you mean, you can't do it? Says, I can't do it, I can't take those words out. Here, so, come, I'll show you. So I went to look, and in fact, he was correct, right? What you saw was this huge glob of sound, and what would happen is you couldn't take a word out, right? Partly because what you get is you get this blurring of these word boundaries, so when you look at the blurry boundary and you look at the divide between blurry and boundary what you find is that e in there it has a little bit of b in it. So it would be like e e ba ba right b b and the b on this side also has a little bit of e in it so it'd be like eb eb eb eb eb right so it actually changes the b you hear after blurry boundary is different than the b you hear after big boundary. So the b you hear after big boundary is different than the b you hear after blurry boundary. And you say, no it's not, it's boundary. Well, that's the magic of our language system that's able to do that, but in fact, there's no divide and there's a little bit of e in that b or a little bit of g in that b, and that all changes, right? Depending on what words are being combined and what sounds are being combined, that's called coarticulation. So in fact, when you look at a spectrogram, you see no clear, defined boundaries, unless you wanted to speak like a robot, right? In which case I, I'll stop because as you see that's very artificial, nobody speaks that way naturally so natural speech has all of this core articulation, it has all of this information basically stuck together. And our auditory system is able to take that information apart, break it up into sounds, break it up into words quite well in a native language. Work by Jenny Saffran at the University of Wisconsin and her colleagues has looked at actually how kids and adults separate these boundaries. So what they did is they in fact created sets of sounds, these are nonsense sounds, and they varied the probability of one sound occurring either within the word or at the boundary of a word, right? And they were very extreme in using this type of probability to determine what words were, but interestingly, they found that in fact children could figure out these boundaries quite well, and so they could actually figure out calculating the probability of what sounds go either within the word or between words where those word boundaries were occurring. This is kind of a statistical type of processing that we do when we process language. So these types of probabilities are, are used in language in other ways. So for example, my daughter Camille who is now nine years of age, but when she was about four years of age, she asked her mom, mom, am I being have? Now, you could think, well, why would she do that? Well, of course, she was asked to be nice, be good be fast. So she assumed that have was like good and nice, an adjective. And so she said, Mom, am I being have? And that was a good way to process given how often she had heard be with an adjective.
