Thursday, May 17, 2007

The Language Instinct (paca)

Through a long chain, believe it or not related to Jerry Falwell's death, I ended up attempting to explain Noam Chomsky's importance to linguistics and psychology on The Moderate Voice blog. Near the end of that chain, I stated that I would explain the idea of the language instinct, which is one of Chomsky's main contributions to cognitive science, on my own blog if people were interested. The term "the language instinct" was made most famous by the cognitive scientist Steven Pinker, whose book of that name was a NYT best seller. However, it is Chomsky who put forth the idea starting in the 60s that language is a biological organ like the brain or visual system. What could this possibly mean?

(Note: Instead of doing a thorough article with references and links, I'm largely going to write this off the top of my head. Also, I am going to do my best to only present some qualifications to the language instinct argument, instead of really considering it critically. The purpose of this essay is largely to establish what the theory is, not assess it.)

To understand why people think there is a language instinct, it might be best to state first what language is not. Language is not just a collection of words and sentences which people memorize through hearing them. Why must this be true? Because it's very likely that I have never in my entire life written many of the sentences I am writing right now. Moreover, I very likely have never heard all of these sentences in my life, and the same goes for you, the reader. And yet, here we all are, understanding each sentence with ease. In fact, language is so easy to us, we don't even realize usually that there's much to it. But the truth is that no computer in the world yet can really do with language what we are all doing unconsciously. In fact, infants who don't seem to do much other than pass liquids in and out of various orifices, cry, and stare can surpass our best attempts at computer language in certain ways.

But if we've never heard these sentences before, how in the world do we understand them? Clearly, it's because we understand how English works. Whatever that means. There's some sort of pattern or structure to English, which we all possess, and which lets us come up with an infinite set of utterances. This structure of a language is typically called its grammar or syntax. Now, sometimes a grammar is defined as any type of linguistic structure, be it the meaning of sentences, the sound patterns, and more. More often grammar is just about acceptable sentence structures and gets explained in terms of nouns, verbs, subjects, objects, etc. Since all of us reading this are able to understand it, it would seem that we all have some sort of mental grammar in our heads that we carry around.

This probably all seems really obvious and trivial, but it's not. In fact, arguing that this is the case was a revolution in the field of linguistics that was started in the late 50s when Chomsky reviewed the behaviorist B.F. Skinner's book "Verbal Behavior". With Skinner, language was treated like all other behaviors. People build some sort of associations between memorized words and the job of scientists who want to understand language is to just explain their behavior with words. Chomsky blew all that up by arguing that in fact the study of language is the study of people's minds. The job of linguistics is to figure out what the hell this structure, this grammar, that we all share is. And then Chomsky went on to make a proposal of what a linguistic grammar might be, called transformational grammar at the time, and refining and improving these models has been considered the central task of linguistics ever since.

So now we have some sort of description of a grammar, but we have a very obvious question. Where does it come from? Is there a unique English grammar and a unique Hindi grammar and a unique Japanese grammar, or is there any merit to the idea of a grammar for all human language?

The obvious solution is to say that each language has its own rules and that children just learn these rules by hanging out with adults. There's some good evidence for this as well, besides being just a good first guess. First up, it's clear that there is no genetic/ethnic relationship between a child and their language. You take any healthy child from anywhere in the world and put them in the right language environment, and, presto, they speak the language in a few years with no problem. Moreover, parts of language are clearly random and learned. There's no particular reason the word for dog in English is begun by putting the tongue up and behind the teeth, while in French it goes back further, and in Spanish is done by putting the lips together, and in Chinese is done by sticking the back of the tongue to the roof of the mouth. People must learn this stuff. There are unfortunately cases where children are not exposed to language, such as when they are abused and locked in closets, and they do not learn language in these cases.

But there are a lot of problems and limitations with this account as well. One set of limitations is that it doesn't explain so-called language universals, which means that languages aren't as random as they often seem. Let's consider basic word order of a sentence. The basic English sentence word order has the subject first, then the verb, then the object. Chinese does the same. So does French. A lot of languages like to put the subject first, then the object, and then the verb at the end. Japanese and German and Korean do this. And there are a lot of languages that start with the verb first, then the subject, and then the object. In sum, we often see SVO, SOV, and VSO word orders. But notice that the subject is always before the object. It's really rare to see the object first. Why would that be the case if language is just random things you happen to hear around you?

There are a lot of other peculiarities that are more persuasive. In English we stick our adjectives before our nouns. "Red ball". We also put our possessive phrases before what they modify. "Paca's essay". Also, in prepositional phrases, we stick the preposition before the noun, "on the chair". It turns out that these things often all go together. If a language puts its objects after its verbs, it will do everything else that seems completely unrelated just like this. But if you find a language that puts the verb after the object, it all flips so that adjectives and prepositions and possessors all go after the noun they modify. If language is just learned patterns, why wouldn't things be a lot more mixed up than this?

All of this however just hints that there is more to human language than random pattern learning, but it doesn't point particularly to a biological factor. Most of the evidence for language as human biology comes from how language is learned - or not learned as it will be argued. First up, every single healthy child becomes fluent in their language. Not every child learns to play the piano as well as others or play baseball. But all speakers of a particular language dialect seem to have the same basic grammar in their heads. Also, there seems to be something of a biological clock in when you can learn language. If you remember the child locked in a closet, when they are rescued, they don't seem to ever be able to learn a language's grammar at all. More every day examples come from second language learning. If you start a language as a very young child, you will learn it easily. Many adults only seem to be able to learn a new language with immense pain and still never become truly fluent. It's kind of like there's a biological clock that gets turned off around puberty.

A lot of evidence for language as biology comes from signed languages as well. Many deaf children can grow up in environments where the signed language they hear is inconsistent or sporadic. The most common reason for this is simply that you have hearing parents with a deaf child who also lacks constant contact with a native signing community. So the parents are there signing really bad ASL, as bad as my Chinese, to their child, but the child's sign language usually ends up better than the parents. The child has a way of ignoring all the screw ups she sees from her parents, and instead making the structure of the language much more regular, much more grammatical. In other words, the child, in a sense, is making up a grammar that she's never fully seen. The most extreme form of this is a case of signed language in Nicaragua. The children were put together in the late 70s in a school for the deaf, and there were no signers around to teach them anything in sign. And yet the children started to invent a language it seems, creating a grammar from scratch. That's completely amazing, isn't it? And the languages that these isolated children come up with isn't weird. It ends up being just like other human languages - languages they've never seen or heard.

The final and most important argument for a language instinct comes from normal, everyday language learning that occurs in the rich linguistic environment that most children grow up with. The problem is that speakers of a language all seem to end up with the same basic grammar of their language, and yet there is no way to choose this grammar from all the possible grammars that fit the sentences they are hearing. In other words, a child hears sentence A, and grammar A could account for it, and so could grammar B, and grammar C, and so on. And there's no clear way when you look at the sentences children hear to know that A is right, where B and C are wrong, and yet we all end up believing A. How is that possible? Moreover, children make a whole bunch of mistakes while learning their language, but there are some they never ever make. Mistakes that should be possible misunderstandings, but in fact never occur.

The classic example of this, from Chomsky, is called structure dependence and it has to do with forming questions. To form a question from the sentence, "Paca is boring me," you take the verb and move it to the front: "Is Paca boring me?" But remember the child is trying to learn the grammar, the pattern, not just one example. So they make a guess at the underlying rule, perhaps they guess "take the first verb and move it to the front." That's a good guess and it works most of the time. But then they want to make a question out of "Paca who is long-winded is boring me." If they follow their guess, they should say, "Is Paca who long-winded is boring me?"

But kids never do this. They don't try this out and then give it up when people look at them crazy. They never do it at all. But if they are just trying out possible grammars by listening to their language, some children should give this a try. But they don't. Why not?

Chomsky's solution to all this is that large components of language are not learned at all. They are part of our biology, just like the design of our eyes, and the way our hormones affect emotions, and everything else. Kids don't try out the weird "first verb" idea, because part of human biology is that sentences have structure with main verbs. And sign language children with bad linguistic input come up with a normal human grammar on their own, because the grammar is inborn in them. And that whole host of things that can come before or after the nouns always seem to follow the same pattern because a single biological parameter - right or left - controls all of them at once. Language then is only partly learned. You still have to learn all the words of your language, but as for the grammar, you already know what is possible in any human language, since we all share the same basic biology, and you only have to discover which of this very basic set is the one you will use. You don't have to start from scratch considering every logical possibility. Finally, like other biological organs, it grows and matures at a set rate. If you are in a linguistic environment while your biology is ready then you will learn like a native speaker. But after that you can only learn with trouble and perhaps never learn fully.

It is in this since that language is a human instinct like a beaver building a dam or a baby kangaroo who crawls blindly into its mother's pouch.

Is it right?

As I spelled it out here, probably no. I think that Chomsky is absolutely right that our biology determines the languages we can learn and how we learn. But it seems patently false that our genes spell out phrase structure or the rules on pronouns, which are indeed part of Chomsky's biological endowment until about 2004. One of the main problems is that it's not clear we have a very good notion of what a human grammar is like or if we all really share the same one, and so it's not much more than speculation on whether or not we can learn it.

3 comments:

Anonymous said...

good article. You can refer to www.about.com

Robin S. said...

This is fascinating stuff.
I need to check in here more often.

pacatrue said...

Thanks, robin s. I hope you do.