2005 LECTURE SERIES

How Children Shape Languages: Language Acquisition and Emergence

Dr. Elissa L. Newport
University of Rochester
Tuesday, Feburary 1, 2005

I’m going to speak this evening about the problem of language acquisition and some work we’ve been doing in that field. There are a couple of points that I want to make.  One is that children are remarkable language learners and probably all of you who have children or have been children are aware of that. 
Perhaps most remarkable is that—what I’m going to talk about is that they don’t always learn what they’re exposed to.  Children, in fact, we think may be a prime force in changing and forming languages. That’s the topic that I really want to focus on tonight.

What I’m going to talk about is something called creolization, the formation, and change, how young languages get formed, and how they change. Evidence that suggests that children are responsible for this process, that who it is that forms languages, who it is that makes languages develop and emerge is really children, the children that are acquiring the language that makes it come about.
I’m going to talk about this evidence from some unusual cases of natural language acquisition that we’ve been working on in my lab and also some experiments that we’ve been doing in our lab that look closely at children versus adults learning languages.

I want to start out by talking about something called Pidgins and Creoles.  You may be familiar with these terms.  You may have heard Pidgin English or Hawaiian Creole, but you may not know the technical definition of those terms and they are relevant for talking about how young languages form.
A Pidgin is a language that forms among adults who share no language in common with one another. In many parts of the world there are times where adults need to interact with one another, but the adults may not have any native language that they share.  Commonly, this happens in regions of the world where there are lots of different languages being spoken. The adults may need to associate with one another for business or for social interactions. They don’t share a native language.

Now in regions of the world where there are only two languages, typically people just become bilingual and learn the other language.  But in regions of the world where there are 40 languages, which is what happens in Papua, New Guinea, for example, people don’t learn all of each other’s languages.  They instead develop what’s called a lingua franca or some method of communication without knowing each other’s full languages. 

Languages like this that are developed among adults are called Pidgins.  Pidgins are defined as languages that are somewhat ad hoc and have no native speakers.
One of the characteristics of Pidgins is that they are grammatically very limited.  They’re typically very simple. Pidgin languages will typically have some basic nouns and verbs, will have some simple word order that they use for communicating, but nothing very complicated. They’re not full, complicated, rich languages.

What’s interesting about them is that from them develops something called a Creole. A Creole—you’ve probably heard of Creole cooking—this is Creole languages.  Hawaiian Creole English is an example.
A Creole is what emerges when Pidgin speakers marry and the Pidgin becomes the native language of the child. A Creole language is a language that is a Pidgin that’s acquired native speakers.  When children learn a Pidgin, even though the Pidgin itself is very limited grammatically, when children learn it as their native language it becomes grammatically more complicated and regular. As soon as you introduce a generation of children into the process languages start to get fancier.

One of the things that’s interesting in our field is why does that happen.  Once you have a Creole language, then subsequent generations start to expand languages. Pretty soon you have something as complicated as the regular languages of the world.

There are various accounts in the field for how languages develop this kind of complexity. I’m going to give you some examples.  But one account is well, maybe it’s because the people who are speaking the Creole, the children are just borrowing more complicated structures from the 40 languages that surround them as they’re learning the Pidgin. 

A second account is called functional expansion.  Maybe languages get more complicated because when it’s really your native language, as it is for the child, there’s a lot more work you need to do with the language.  There are more functions that you need to perform.  And so because you use it for more purposes, it just naturally gets more complicated.

But the account that I want to focus on tonight is what I’ve been referring to that the reason languages become more complicated as children learn them as their native language is because this is just a natural part of language acquisition.  It’s really the children learning the language that make it get more complicated and more elaborated.

So it’s the evidence for that notion that I want to talk about.

I want to illustrate this by an ongoing piece of work that’s been done partly starting in my lab and partly continuing independently by someone who is a former post doc of mine, Annie Senghas. This is a language that is undergoing the sort of process right now in Nicaragua. 

Most of the time when Creole languages have formed no one’s been around to watch how they form. Most of the work on this topic comes from looking retrospectively and trying to figure out without really good observations exactly what forces make the language expand.

In this particular case there’s a very unusual opportunity where people have actually been there, linguists have been there studying the language as it expands and have provided some special insights into what kinds of things make languages get more complicated. 

I wanted to tell you a little bit about the nature of this particular language.
This picture is illustrating the way in which this particular language began.  Nicaraguan Sign Language is a sign language that’s been developing among deaf people in Nicaragua.  And at the origins, as best anyone can tell, deaf people in Nicaragua were generally born into hearing families.  It’s very common for a deaf child to be the only deaf person born into a hearing family. 

Under those circumstances, that’s what these little individual houses are illustrating, the deaf children in Nicaragua, as best anyone can tell, didn’t really interact with each other.  They didn’t go to school. They never met up with each other. Under those circumstances what typically happens is that children develop what’s called a home sign system. This means a deaf child will begin to communicate with their parents by gesture. The family develops some kind of simple gestural communication system that has properties sort of like Pidgin languages that I’ve been describing.

In Nicaragua the interesting thing that happened is about 1980 it was decided—the Nicaraguan government decided to begin public deaf education. At that time they rounded up all the deaf kids in the vicinity of Managua, which is the capital, and began starting a school. 

Now unfortunately the school doesn’t really do a great job of education.  At the beginning it ran an hour a day.  They tended to speak to the kids in spoken Spanish.  And it really wasn’t the school itself, or the educational practices that changed the circumstances in Nicaragua. 

What happened in Nicaragua is simply that all the deaf children who didn’t know each other ran into each other at school. That has happened in various places around the world.  When that happens, as soon as deaf children become in contact with one another, you start to see the emergence of something that becomes a community language.  And then start to develop some properties of the language—of the conventional language system.

That’s what we’ve called the first cohort.  The first group of children, older and younger children who are brought together by the school and started to interact with each other and started to form some kind of common sign language.

Then that process has continued until today what’s actually happened is that there have been continuous new children coming in as new deaf children are born in the vicinity or new children are located.  They have joined the school. There’s a continual addition of the community, children graduating, getting older and graduating and joining the community and learning the sign language from other children.
Today we have sort of arbitrarily grouped them and called the people who have come in more recently, after the mid-80s, the second cohort. 

Now this is just another kind of representation that you have cohort one as the people who joined the community in the early 80s.  And cohort two, we again have arbitrarily grouped as the cohort that began in the mid 1980s. 

What’s of interest is that the differences in time and in the generation or cohort of the children have really created striking differences in the nature of the language.  And this is what—an example of what’s called creolization. Again, because people have been there to observe this happening, we have some insight into the causes or the processes by which this has changed.

I wanted to illustrate a little bit of this to you from Annie Senghas’ work.  First I just want to show you one kind of example that you can see in a small movie of the difference between, in this case, what’s being illustrated at the top is a hearing person’s gesture for a complex event.  And at the bottom—I’m going to show you this in a film and then a slide—a second cohort child who is using Nicaraguan Sign Language rather than gesture for expressing the same kind of event.  Both of these people are being shown pictures of an object that wobbles along a path.  And they are just asked to describe it.

What I want to emphasize is when you see gesture, you’ll see that this person does the whole thing in a kind of mime.  He takes his hand and moves it back and forth as the object would travel along the path. When you look at the child who’s beginning to use this as a language, what you see is that the pieces of the event start to be pulled apart and sequenced more like a sentence. You start to see a sequential sort of representation in gestures of something that in the actual world is happening all at the same time. There’s a characteristic feature of languages that instead of acting out something that’s all simultaneous, we pull it apart and express the parts of the event in separate words in some kind of sequence. 
Let me just show you this.  First you’re going to see the gesturer doing everything simultaneously, doing something that looks like this. 
[Video]

Then in contrast, this is the child who is doing it as part of Nicaraguan sign language. He’s actually talking about the same event.  But what you’re going to see is that first he describes the moving around and then the path.  He’s talking about the same event with an object that’s actually moving back and forth while it moves forward. In a language one tends to pull these pieces apart and put them in sequence.
[Video]

He does.  And that’s, even though they’re very similar, this is just the beginning hint that this is really like a language expresses it instead of like a gesture would express the same notion.
What I want to show you now is some quantitative data that Annie Senghas has collected to show other ways in which the language has changed. This is the figure I showed you already. I just wanted to remind you that cohort one, which you’re going to see on the next slide are the people who started in the community in the early beginning 1980. Cohort one is like the second generation of children to whom this language has been passed on who entered the community later and who learned the language from cohort one.

To remind you, this is a situation in which deaf children are entering the community and learning the language from a previous generation of children, unlike in most of our languages it’s not a situation in which children are learning from their parents. They’re learning—cohort two is learning from cohort one.
This is a figure that is a representation of the speed of the language. One of the things that happens as languages get more complicated and start to form is that they get faster.  If you look at Creole speakers, speakers who have a more well developed language, the younger children, the people who have learned it later as children speak faster than the adults from whom they learned the language. 

This happens in the Nicaraguan Sign Language data too.  This is a little complicated.  Let me show you how to read this figure.  There are actually two variables that are being looked at here so the set of bars on the left are people who learn the language who joined the signing community who went to school when they were young children, early in their lives before they were about six. 

Here are kids who joined the community when they were six to ten, and over on the right are people who joined the community when they were older than ten. The variable being displayed here is well, how old were you when you were first got exposed to the language?

And the other variable, the white versus the black bars are the two cohorts.  Were you in the first cohort or were you in the second cohort?

There are facts of both of these variables that you can see.  The measure here is how many more themes or bits of meaning per minute people are producing.   It’s just a measure of speed.  The higher the rate, the faster people are signing.

What you can see here is that there are effects of both of these.  So in general, the younger you were when you got exposed to the language, the faster you sign. 
Now you may think well, yeah, kids talk really fast.  But that’s actually not the effect here.  People who began learning the language as children, their whole lives speak faster, sign faster than people who began learning the language late in life. This is a reflection of their fluency in the language.
People who learned the language after age 10 always sign slowly and have less fluency in the language.
You can see that the speed drops as you get to people who started being exposed to the language later and later in life.

In addition there’s another effect, which is if you began early or in your middle childhood, the second cohort got even faster than the first. Each of these groups, the black line, the second cohort was learning to sign from the first cohort shown in white. What’s happening is that in each case you’re getting faster.  These kids are learning from medium fast children before them.  But they’re signing even faster. The language is kind of accelerating in the fluency that people show as a function of having been exposed to the people before them.

That effect doesn’t appear in the late learners.  Late learners, because they were learning it late in life are just going to not be so good no matter what.

This is a figure that looks exactly the same. It’s actually measuring something different.  This is measuring what’s called spatial modulations.  In sign languages the way you express who’s doing what to whom is typically in many sign languages you’ll set up people in space.  You may point to a location over here for one person and point to another location for another.  And there’s movement from one location to another indicating who did what to whom.  And these are called spatial modulations.
That’s an indication of a more complicated language. Again, you can see that the earlier you learn the language, the more spatial modulations, the more manipulations of space people make. 
Again, there’s a cohort effect where the second cohort gets more complicated and uses space more than the first. 

In both of these cases you have two phenomena.  One is an age of learning effect. The other is the next generation gets better than its predecessor or the generation it learned from.
Again, there are two phenomena here.  One is called a critical period phenomenon.  There seems to be a prime time in life early in life where people learn languages really best. If you’ve tried language learning lately, you probably are aware that that’s true.  Children acquire languages better than adults.
In addition, there’s this second phenomenon, which is children go beyond their input.  If they get input to somebody who is not so great, they get better than their input. Together that produces a process in which young languages, with children being the transmitters, get more and more complicated as they go.
I want to show you other data for both of these two things I’m going to talk about.  One is the critical period phenomenon.  Again, you’ve probably all noticed this.  But it’s not well studied when you look around the world.  Is it really the case that kids are better at learning languages than their parents?  They look better.  If you’ve moved to a new country with your children, typically you’ll acquire the language not so well.  And your children will be off and running.

When you look around the world it looks like kids are better language learners. There have been a number of other accounts.  Children might not be as good at languages—children might be better at languages because they get more exposure or because they try harder.  Adults may not be exposed to the new language as much.

This is what a critical period looks like if it’s the case that there really is a prime period in life for learning any behavior, then you’ll see that over age of exposure you’ll turn out best if you learn at a particular period.  And at another period in life the same exposure will not result in such a great performance.
One can ask does this really happen in the acquisition of languages, if you try to control all the relevant factors, everybody’s trying hard, everybody’s got the same amount of exposure, is it really the case that people get better at learning languages at particular times. 

We have done some work where what we’ve done is to vary the age of arrival of people in the United States.  This is now a study of hearing people who came to the United States at different ages and learned their language for a fixed period of time.  These are all people who learned their language for 10 years before we studied them.  And then we measured some aspect of their proficiency in English as their first language.  And then we asked do people really show an advantage for learning early in life as opposed to late.

What we did was in this study we took people whose first language was Chinese or Korean. We gave them a judgment task.  We gave them sentences like the boy is at the door versus boy is at the door. We asked them to perform the task and just say whether the sentences sounded correct or not correct when they were exposed to them.

These data are the data that we got when we did this with people who moved to the United States at various ages.  So over here you see people who were all native speakers of English. They were asked to judge on this task. You can see that they do very well.  That’s sort of how you define the language. As people moved to the United States older and older, they did more and more poorly in the task.
All of these people were tested about 10 years after they arrived. They were all matched for the length of time that they had been here.  But you can see that there is this dramatic effect of the age that they were actually exposed to the language.

Then there’s a falling off or a plateau where people who arrived in adulthood are pretty much all the same. This period of time is about the time over which you find brain maturation in children.  As the brain matures, you find a falling off of the ability to learn a language.
This picture is the same thing. Each of these dots is a person. This is different people who arrived in the United States at different ages. You can see that in general there is a decline where the average performance in the language falls off.

The other thing I wanted to point out is that you’ll also see this increase in variability where, if you’re a child everybody’s up here and really great in the language. As people continue to be older you find much more individual difference. Some people are very good.  But the average level of performance is not so good.

We also have data on first language acquisition with deaf people who are learning American Sign Language at various ages. You find exactly the same kind of effect if you look at people who are exposed to the language from birth, which is what we call native speakers.  They’re all really good at learning the language.  This is a variety of tasks.  It doesn’t really matter what kind of task you look at.  These are a mixture of both production and comprehension tasks. People who are exposed to the language early in life turn out to be really good at language. As they get older in their first exposure to the language, they get to be less and less proficient on the average.

Now this study was done with people who had been signing every day for about 50 or more years. This is a huge amount of exposure to the language. Nonetheless, there’s still a major permanent effect of not being exposed to the language until you’re an adult.

These data say there is for even full languages, there is an age effect or a critical period.  There’s something affecting the success in acquiring a language.  Children are much better at learning a language than adults are, even when the language is a full-blown language.

These are—now I’m going to turn to data for the second part of what I said.  If you’re a child and you are learning the language early in life, you learn the language best—sorry, you learn the language better than the input that you’re exposed to. 

If you’re a child, and you’re learning the language early in life, you can be exposed to imperfect input.  The people that you learn from don’t have to be completely proficient speakers of the language.  You can still get better than the people that you learn from.

This is a study that I’ve done of a set of children—I’m now going back to the sign language case again—a set of children who are learning to sign American Sign Language from people who are their parents, but where the parents are not fully proficient in the language.

The red bar here is the child we studied for quite some time who we call Simon. At the time we tested him here he was nine years old. Let me explain the circumstances in which Simon learned his language. 
Simon learned to sign from his Mom and Dad. The mother and dad here are deaf people who learned to sign themselves when they were adults. 

If we go back to this figure, Simon is a native speaker of the language.  He’s exposed to the language from birth. His parents come from this late learning group.  His parents were first exposed to American Sign Language when they were teenagers.  They spent their lives being trained orally.  And they learned to sign when they were 15 and 16 years old.

You can see from this graph that his parents would not be totally proficient in the language.
What happens if you’re in the group down here is roughly you get things right.  You get lots of things right, as you probably know if you were a late learner of a language.  But you also make lots of mistakes. You’re very inconsistent in the way you use the structures of the language. This is characteristic of late learners of spoken languages.  It’s also characteristic of late learners of sign languages.
Simon is coming from this native speaking group.  He’s exposed to American Sign Language from birth.  But the people that he’s learning it from are late learners and they use the language middling well. They make lots of errors.

The question we asked in this study was well, what happens to Simon under these circumstances?  He has the advantage that he’s a child.  But he has the disadvantage that the people he’s learning the language from use it inconsistently. What does his signing turn out to look like?
These data come from a study of a Simon, as I said, when he was nine.  We followed him longitudinally. We looked at his signing for his whole life. Looked at how his development worked. These are pretty much when he’s done learning the language and he’s quite a successful speaker.

What we did was to show him a set of events that involved motion.  And we asked him to describe them. Then we scored him compared with native ASL, American Sign Language, to see how he did.
You can see this axis shows that he’s doing about 90% of the time exactly what he should be in perfect ASL. He compares extremely favorably to kids over here who are also native speakers of American Sign Language, but whose parents were perfect proficient signers. You can see that he’s about as good as they are.  In fact, if anything he’s a little better than they are.

However, over here are his Mom and his Dad.  His mother and his Dad make lots of mistakes.  His Mom is correct about 75% of the time so she’s not a bad signer.  This is true for many people who learn the language late in life.  They’re pretty good.  But she’s making lots of mistakes.  She’s making mistakes about 25% of the time. 

The point here is that Simon is exposed only to his parents.  He doesn’t actually know anybody who signs better than his parents.  They’re making a fair number of mistakes.  But Simon notice is substantially better than they are. This is the second part of what I was talking about for if children are exposed young in life, but exposed to people who are not perfect where the input is not consistent, the child can still get better than their own parents.  They can find the part of what their parents do that is consistent and improve upon it and make the language that they use more consistent and more regular.
That’s a phenomenon that we’ve seen in this one child that we’ve followed. We’ve also seen this same phenomenon in other children that we’ve studied. 

This is a child called Susan. You see the same phenomenon in Susan.  Susan is about nine when we studied her also. You can see this is Susan tested on the same task.  Susan is really doing well.  She’s scoring about 90% correct on this task.  She’s scoring exactly as well as a set of children whose parents are native speakers.  But here is Susan’s Mom and Dad. Susan’s Mom and Dad are much worse than Susan. 
Again, these are the only people that Susan knows who know this language. She’s learning the language from her Mom and Dad who are making a lot of mistakes. Susan is surpassing them.  She is finding the part of what they do that is the most consistent.  And she is more consistent, better in the language than they are.

Here’s Stewart who is about the same age. Stewart’s Mom and Dad also aren’t such great signers.  In this case, if we go back to Susan, you can see that her Mom’s pretty good and her Dad is really bad.  And she is better than both of them. 

In Stewart’s case, the Dad is a pretty good signer.  He is not doing too badly.  Her mother is really not so good.  But Stewart—sorry, his mother is really not so good, but Stewart is better than both of them again. Again, just as good as kids who have native exposure, whose parents are native speakers of the language.
Sarah is our most surprising case.  This is Sarah’s mother.  Sarah has a single mother and no Dad. Sarah’s mother is really not a good signer.  Sarah’s Mom just learned to sign when she found out she had a deaf child.  She’s doing her best. She’s really not such a great signer.  She’s getting about 15% correct on our task. That little 15% she gets right is pretty consistent.  And then she’s making a huge number of errors that are scattered all over the place. Sarah seems to be able to focus just on the part that her Mom does consistently, do that, and become much better than her Mom.
Again, even though she doesn’t know anybody who signs the language very well, she is substantially better than they are.

The phenomenon here in all of these kids, just to bring this a little closer to something that you might be more familiar with, imagine that you are exposed to say Chinese.  Suppose that your parents spoke Chinese to you at home, and they were native speakers of English, and their Chinese wasn’t so good.  They were often right in their Chinese, but making lots of mistakes. 

The phenomenon here is you could still nonetheless sort of ferret out what’s the most consistent about what your parents did and manage to acquire that and get much better than your parents.
What all of these children are showing is that they can get better than the kind of input that they receive.
Now what we did next was to ask whether we could investigate through some experiments whether this was really something that all learners do or whether this was special to kids. 
What happens, in fact, when learners get exposed to fairly inconsistent input?  Can anybody get better than their input, and figure out what’s consistent and improve upon it or is this only a property of children.

The first experiment we did was with adult learners.  We made little miniature languages in the lab. We made them partially inconsistent. We asked whether adult learners would acquire languages and get better than their inconsistent input.

What do adult learners do when they’re exposed to inconsistent input?

Now, I’m going to show you in a minute how we make a little miniature language and how we expose people to this in the lab.  But we had adult subjects.  These were college undergraduates who came into the lab who we paid to learn these little miniature languages.  They came in for a number of sessions.  We made small languages. They came in for five to eight sessions and learned the languages over that time. 

They were little spoken languages.  I’ll show you in a minute what they looked like.  We showed them little videotapes of events. Then they heard sentences in this made up language. They learned to speak the language.

Then at the end of the time that we exposed them, they had to say some new sentences.  The language was what’s called a VSO language.  It’s a verb first, and then the subject, and then the object. It would be something like if in English you said hit John ball. The reason we did that was just to make the language differently organized.  Languages of the world do have that kind of structure.  But English doesn’t. This was something that would be different than the language that they already knew.
What we did was to add one little bit of the language that was inconsistently used.  There was a little form called a determiner, which is like the word the, or, a, in English, which goes with nouns. In English we say the boy.  In this language you would say in the nonsense word something like boy the. The determiner was used inconsistently.

This was replicating something like when Simon’s parents inconsistently use a structure in American Sign Language. We varied the amount of inconsistency. Some subjects got exposed to the little determiners used perfectly all the time that they should be there. That’s the 100% consistent case.  Some subjects got the little determiners there 75% of the time and— what we wanted to see was well, how do learners who are adults do under these circumstances.

As I said, they got exposed by videotapes. They had six to eight exposure sessions. All together they had about three to four hours of exposure to the language.  But these were little tiny languages.

Let me show you one of these miniature languages.  What you’re going to see is a little videotape.  And then you’re going to hear someone speaking these sentences. You can hear—this is a language we made up.  But it kind of sounds like a real language.
[Video]
There will be a test.

[Video]
Okay.  Ready for the test? 

The little words we’re interested in but in general people had no trouble learning this language. On almost everything about the language, the words were used consistently.  They had a consistent word order.  And everybody acquired this language perfectly in all the conditions. 

The little things that we were especially interested in, see these little—the po and the ca.  These were little things that sometimes occurred with nouns and sometimes were omitted. 

What we were interested in was if those words were used consistently, we knew they would be acquired well.  What if they were used inconsistently like Simon’s parents would inconsistently use some of the grammar of American Sign Language?  And again, this is adult learners who we were testing in this study.
The way we tested this was a sentence completion task.  We gave them new videotapes, events that they had never seen before. Then we gave them the first word of the sentence. Here’s prog, that’s the verb in this language.  And then they had to say the rest of the sentence. The person had to speak the language saying a sentence that would describe this event that they saw.

Again, I’m going to show you and you get hints here about what the right answer is.  And I’m going to show you a subject who did this pretty well.
So they see an event. 
[Video]
Fluent.

[Video]
You get the right answers over here. 

[Video]
He was really good.  He was in the 100% condition.

Let me show you what happens when you have adult learners and they get perfect input versus inconsistent input. What you’re going to see plotted here is the four groups.  Remember I said that there were four groups in terms of how often these little determiner words were used.  Over here the determiners were used 100% of the time.  That is every noun had one of these little words, po or ca with it.  And here you had only 75% of the time, 60% of the time, or 45% of the time. 

The input, this dotted line is the input that they received, the language that they were exposed to, and how often these determiners were used. 

What we were interested in was did they like Simon go up here, did they all use the form 90% of the time getting more consistent and better than their input or not.

Here is actually what you get in the adult speakers.  It’s kind of remarkable.  Instead of making things more regular, they actually look incredibly like their input.  If they hear the little determiners 60% of the time, they’re almost exactly the same likelihood to use the determiners themselves. 
Notice nobody’s up here.  Nobody’s actually being more consistent than the input that they received.  They’re actually really matching the inconsistencies.  They’re getting a little bit worse. This is typical of adult learners.  If you’re exposed to things that are not so perfect, you actually make them a little worse.
Then what we did was to ask well, what happens if you compared this to children. What we found was a kind of surprising result.  We ran this same experiment.  We made the language a little bit smaller and simpler.  But basically we ran the same language with six to seven year old children.  And then we reran some adults under the same circumstances as the children.

This was live presentation because the kids didn’t learn the language so well from videotape.  You may be surprised, even though kids like to watch video; it’s not so easy to learn a language by that method. 
We did a live presentation. We just did two conditions.  We had these determiners used perfectly that is 100% of the time and 60% of the time. Here’s what you see if you just look at how frequently they used the main determiner that was the po or ca that was supposed to be used.
Adults in the 100% condition do it 100% of the time.  In the 60% condition, they do it a little less than 60% of the time. For adults this is like what we saw before.

Kids don’t look so good by this measure.  Notice kids look like they’re blah; they’re not so great.
But there actually turned out to be something that we discovered when we looked at children more carefully that was very different from the adult performance.  If you score somewhat differently, and you ask well, how consistent are children.  If you look carefully at their production, is there some rule that governs what they’re doing. 

It turns out that the children almost always in both groups have some consistency. I’m going to give you some examples.  All the kids are forming some kind of a rule, whereas the adults are not.  In the perfect condition they use the determiners perfectly.  But in the 60% condition they’re still very probabilistic. 
Here are some examples of what the kids did.  They sometimes used this main determiner form all the time with nouns in transitive sentences, that is things like the boy hit the ball.  But never in intransitive sentences like the pin fell over.

They had a perfect distinction.  All the time they used the determiner in one kind of sentence, never in another.

Now this contrast was not at all what their input had.  Their input just did it 60% of the time at random. What kids did was found some kind of rule.  And imposed that really regular kind of rule on their input.  Or they used their main determiner always for object nouns and never for subject nouns.
One after the other the kids formed some kind of consistent pattern.  They found some kind of consistency and imposed that on what they did.

This is over a bunch of different conditions that I’m not going to describe, just a scoring of looking at the kids versus the adults.  How many times were kids perfectly consistent in how they used these inconsistently used determiners.  And how many times were adults. 

We’ve categorized them so that the light blue color is if you had a perfect rule, you were perfectly consistent internally, and you can see that adults—virtually none of the adults were perfectly consistent if they got inconsistent input.  Almost all of them used their form inconsistently like their input did.  But kids were really consistent and always formed some kind of a rule when they were exposed to inconsistent input.

What we seem to be seeing in all of this work, let me try to pull these findings together, is that first of all, as I said, in the first part, if you’re exposed to a regular language, it looks in our data like kids are really better language learners than adults.

In this more recent kind of work on creolization, and when we look at kids learning languages in the lab, there seems to be something extra special about kids.  That children may also be the force that makes languages take their shape and get more regular, that even if they’re exposed to inconsistent input, they seem to go beyond it and to make something more complicated and more regular. 

If we go back to the Nicaraguan Sign Language case, which I started with, as I said earlier, you look at this first cohort and if you look at the consistencies of the first cohort, they’re showing moderate use of the language.  They are pretty fast.  They start to use forms that are organized in space pretty often. 
Then there comes a next generation of kids who learn the language from the first. The tendency of kids when they come into this situation is even when their predecessors are inconsistent, kids make the language more consistent, more complicated, and more regular.

All of this research seems to suggest that when you look at children, they have some extraordinary gifts in learning language, and that in fact, it may be the process of learning language that makes languages form and get more complicated.

Not only are children really gifted at learning regular languages that are already complicated, but it looks like they’re the force that pushes languages to form and become what they are.

Thank you.