Two Theories of Language Acquisition

The history of child language acquisition goes back to the 1950s. There’s a famous debate between Skinner, who had a behaviorist account of how children learn language, which was then famously challenged by Chomsky, who argued that you can’t learn language just by listening to the language that you hear – there must be some kind of system already in your head, which is processing that language in us, that comes in a specific way to allow you to build a grammar from the very small and fragmentary mass of language that you hear. In one way or another, this debate between Chomsky and supporters, who think that you have to have some innate knowledge of language, some language that you’re born with, and various opponents who think that language can indeed be learned just from what you hear, has never really gone away and has continued right up until the present day.

The Chomskyan idea is that children are born with some knowledge of language already in their heads. Obviously, that knowledge has to be very abstract, because the innate knowledge doesn’t know what language the child is going to be born speaking. Obviously, it would be useless to be born with individual words of English or Russian, or whatever, because it would be no use if you’re there in the wrong language, so it has to be very abstract. So, the idea, for example, is that children are born with grammatical categories, like empty boxes. They’re born knowing that languages contain nouns and verbs, and all they have to do is hear the language around them, and once they recognize words that they hear as a noun or a verb, then they can put them in the right box as it were. This helps the child because they’re born not just with the boxes, but with rules for combining these boxes or these categories into sentences.

One rule that child is born with is that you can combine a noun and a verb to make a very simple sentence, like “John ran” or “Bill danced”. The idea is that you don’t even need to learn to do that, you’re born with the categories and the rule for combining them into sentences. You have to learn which order they come in your particular language, but the idea that you can combine a noun phrase and a verb phrase in that way, the idea in the Chomskyan approach is that you don’t have to learn that, you’re born with it.

The other approach says that you’re not born with this knowledge of language – everything that you know about language is just built from the language that you hear all around you. It’s difficult to tell between them, because, obviously, we can’t look directly in children’s heads and see what’s there. The way I try to answer this question is by running experiments with children.

I’ll just talk about one example of where the two approaches make different predictions. So, questions in English is an area that’s been studied quite a lot, because children spend quite a few hours with questions. An English question should be something like “What do you like?” or “What can she eat?” But English-speaking children often make mistakes where they say things like “What she can eat?” or “What you do like?” A way to compare the two approaches is to see if there’s any patterning to these errors.

According to the Chomskyan approach, these questions are formed by one of these rules I mentioned earlier, these rules that you’re born with. If you can make a question with one WH-word on one person, then you should be able to do it for everyone, you shouldn’t be showing different performance for different ones. Whereas the other approach, the constructivist approach, says that you learn language from what you hear around you. This predicts that children might be very good at asking a question like “What are you doing?”, that they’ve heard an awful lot or any question like “What is [thing] [process]. Whereas the question that they haven’t asked very much like “Why isn’t he doing something?” or “Why don’t they do this?”, they would be much worse with those types of questions.

The Chomskyan approach predicts equal performance with all the different types of questions. The constructivist approach predicts better performance with the question types that you hear more frequently. At least in my studies, that’s what we seem to find. Children are very good at questions they’ve heard a lot like “What’s that?”, and much higher error rates (50% or higher) for some of the question types that they hear more rarely. That’s just one example.

Basically, what I do in my studies is just do this over and over again for different sentence types: for questions, for passive sentences, and also for inflectional morphology (the morphemes, the little markers that go on the ends of words). Like in English you have “play-plays-played” or in Russian “играешь-играю-играете” and so on. Again, the Chomskyan approach would say: “Once you’ve learned the verb and you’ve learned the morpheme, it’s a rule of person together, you should be equally good with all verbs and all morphemes. Whereas the constructivist approach says: “No, you start by learning a whole word form and you should be very good with word forms that you’ve heard frequently and bad with ones that you’ve heard much less frequently”.

A difficult thing about the universal grammar debate is that it means lots of different things to different people or even lots of different things to the same people at different points in time. Sometimes when people mean universal grammar, they are talking about the things I mentioned before, these very specific things like categories that you’re born with and rules for combining them to make sentences. But sometimes people use universal grammar to mean just the ability to learn a language. If you present these arguments to Chomsky and say this means there’s no universal grammar, he says: “No, that’s ridiculous. Universal grammar is what means that a human child can learn language and my cat can’t learn language”. So, by definition a child has a universal grammar. In that level, of course, we can all agree with the idea of universal grammar.

The debate over whether there’s universal grammar is whether we’re born with these kinds of very specific categories and rules, so with whether there can be any conciliation between the two approaches depends on exactly what you mean by a universal grammar. If we’re talking about very general things like a bias to be interested in sounds, to be interested in other people and what they seem to be trying to convey or whatever Chomsky means when he says whatever it is that allows humans to learn language and not cats, or trees, then, of course, we can all agree with the universal grammar. But I can’t see any conciliation between approaches that posit that you are born with an absolute knowledge of categories or that you are not. This seems kind of a black-and-white.

Of course, linguistic studies are a very interesting way of trying to choose between these two approaches, but again it’s difficult to do because the findings can be interrupted either way. I’m running studies that are looking at causation, the language of causation across five different languages. We’re doing English, Hebrew, Quiche, Japanese and Hindi. I’m finding there are some real similarities across these different languages and the way causation works. English – and I think this works the same pretty much in Russian as well – has basically two different ways of talking about causative. One is just a straightforward causative like “The man broke the plate”, so the causation is just already there, a kind of straightforward subject-verb-object sentence. But we’ve also got the second type of causation, where you have to add the verb “make”. So, you can’t say “The joke laughed the man”, you have to say “The joke made the man laugh”. What’s interesting is – I wouldn’t want to say all languages, but a lot of languages, and all of the languages that I’m studying – this project show a very similar pattern. They differ hugely in how they do this, but all of these languages have these two different types of ways of marking cause.

We’ve got the same 60 actions or the same 60 verbs across all these different languages, and we’re looking to see if there are any similarities between the languages, in which verbs prefer the “make-type” causative, and which prefer the other type causative. What we’re finding is there are real similarities here. You can use the performance of English speakers to predict the performance of Hebrew speakers or Mayan speakers because there’s so much similarity across 30 languages.

There are different ways you can interpret that. If you were to come at it from a very Chomskyan perspective, you could say that speakers are born with this knowledge of kind of indirect versus direct causality, and the grammatical structures that make that possible. If you’re coming at it, like I am, from the other direction, you can say that these are products of linguistic evolution. Speakers of all languages find it important to talk about causes and to distinguish more direct causes from less direct causes, and that’s why we’ve evolved similar grammars. We can try and answer these questions with cross-linguistic studies, but it always comes back to the same debate. You can always interpret the data in these two different ways.

So, suppose we meet Martians or someone from a different planet, how would their language be like ours? I guess, from my perspective at least, it’s to do with the extent to which they are similar to us in the things they want to get across. As I mentioned before with the cross-linguistic studies, the reason that English and Hebrew all have these two different types of marking cause is that it seems important to us, for human speakers, to be able to discuss causes. Presumably, this is because if a guy goes around causing people to be dead, then that’s important, that we can share that information. If Martians or these speakers from whatever planet had bodies like ours and they could cause injury to each other or they hid them, or do good things to each other, then I would expect that they would also have different ways of marking things like a cause or who was doing what to whom in their language. But if they would just aim off as gases and they didn’t act upon one another, then, I would think, they wouldn’t have these types of a hand of ways of talking about the cause, because it’s not relevant to their society.

Chomsky’s latest thinking on this is that language evolved for thought, so the language evolved for humans to process their thoughts internally, and the fact that we can use it for communicating with other humans is just a happy accident, a byproduct of that. I would have to say that view is quite controversial even within Chomskyan circles. Chomsky and some of his co-authors have argued that recently, but not all of the people who follow a Chomsky in view would say that. Certainly, for those of us on the other side of the fence language is all about communication, and the reason we have the grammatical structures we do is because it’s important for humans to be able to convey these types of meanings. It all goes back to meaning and to pragmatics on our side.