The Language Instinct: How the Mind Creates Language (49 page)

If children could count on being corrected for making such errors, they could take their chances. But in a world of grammatically oblivious parents, they must be more cautious—if they ever went too far and produced ungrammatical sentences together with the grammatical ones, the world would never tell them they were wrong. They would speak ungrammatically all their lives—though a better way of putting it is that that part of the language, the prohibition against the sentence types that the child was using, would not last beyond a single generation. Thus any no-feedback situation presents a difficult challenge to the design of a learning system, and it is of considerable interest to mathematicians, psychologists, and engineers studying learning in general.

How is the child designed to cope with the problem? A good start would be to build in the basic organization of grammar, so the child would try out only the kinds of generalizations that are possible in the world’s languages. Dead ends like
Who did Sheila see Mary and?
, not grammatical in any language, should not even occur to a child, and indeed, no child (or adult) we know of has ever tried it. But this is not enough, because the child also has to figure out how far to leap in the particular language being acquired, and languages vary: some allow many word orders, some only a few; some allow the causative rule to apply freely, others to only a few kinds of verb. Therefore a well-designed child, when faced with several choices in how far to generalize, should, in general, be consecutive: start with the smallest hypothesis about the language that is consistent with what parents say, then expand it outward as the evidence requires. Studies of children’s language show that by and large that is how they work. For example, children learning English never leap to the conclusion that it is a free-word-order language and speak in all orders like
give doggie paper; give paper doggie, paper doggie give; doggie paper give
, and so on. Logically speaking, though, that would be consistent with what they hear if they were willing to entertain the possibility that their parents were just taciturn speakers of Korean, Russian, or Swedish, where several orders are possible. But children learning Korean, Russian, and Swedish
do
sometimes err on the side of caution and use only one of the orders allowed in the language, pending further evidence.

Furthermore, in cases where children do make errors and recover, their grammars must have some internal checks and balances, so that hearing one kind of sentence can catapult another out of the grammar. For example, if the word-building system is organized so that an irregular form listed in the mental dictionary blocks the application of the corresponding rule, hearing
held
enough times will eventually drive out
holded
.

These general conclusions about language learning are interesting, but we would understand them better if we could trace out what actually happens from moment to moment in children’s minds as sentences come in and they try to distill rules from them. Viewed up close, the problem of learning rules is even harder than it appears from a distance. Imagine a hypothetical child trying to extract patterns from the following sentences, without any innate guidance as to how human grammar works:

At first glance, patterns jump out. Sentences, the child might conclude, consist of three words: the first must be
Jane
, the second either
eats
or
likes
, the third
chicken
or
fish
. With these micro-rules, the child can already generalize beyond the input, to the brand-new sentence
Jane likes chicken
. So far, so good. But let’s say the next two sentences are

The word
might
gets added to the list of words that can appear in second position, and the word
slowly
is added to the list that can appear in third position. But look at the generalizations this would allow:

Bad start. The same ambiguity that bedevils language parsing in the adult bedevils language acquisition in the child. The moral is that the child must couch rules in grammatical categories like noun, verb, and auxiliary, not in actual words. That way,
fish
as a noun and
fish
as a verb would be kept separate, and the child would not adulterate the noun rule with instances of verbs and vice versa.

How might a child assign words into categories like noun and verb? Clearly, their meanings help. In all languages, words for objects and people are nouns or noun phrases, words for actions and change of state are verbs. (As we saw in Chapter 4, the converse is not true—many nouns, like
destruction
, do not refer to objects and people, and many verbs, like
interest
, do not refer to actions or changes of state.) Similarly, words for kinds of paths and places are prepositions, and words for qualities tend to be adjectives. Recall that children’s first words refer to objects, actions, directions, and qualities. This is convenient. If children are willing to guess that words for objects are nouns, words for actions are verbs, and so on, they would have a leg up on the rule-learning problem.

But words are not enough; they must be ordered. Imagine the child trying to figure out what kind of word can occur before the verb
bother
. It can’t be done:

The problem is obvious. There
is
a certain something that must come before the verb
bother
, but that something is not a kind of word; it is a kind of
phrase
, a noun phrase. A noun phrase always contains a head noun, but that noun can be followed by all kinds of stuff. So it is hopeless to try to learn a language by analyzing sentences word by word. The child must look for phrases.

What does it mean to look for phrases? A phrase is a group of words. For a sentence of four words, there are eight possible ways to group the words into phrases: {That} {dog bothers me}; {That dog} {bothers me}; {That} {dog bothers} {me}, and so on. For a sentence of five words, there are sixteen possible ways; for a sentence of six words, thirty-two ways; for a sentence of
n
words, 2
^n-1
—a big number for long sentences. Most of these partitionings would give the child groups of words that would be useless in constructing new sentences, such as
wears bothers
and
cheering too
, but the child, unable to rely on parental feedback, has no way of knowing this. Once again, children cannot attack the language-learning task like a logician free of preconceptions; they need guidance.

This guidance could come from two sources. First, the child could assume that parents’ speech respects the basic design of human phrase structure: phrases contain heads; role-players are grouped with heads in the mini-phrases called X-bars; X-bars are grouped with their modifiers inside X-phrases (noun phrase, verb phrase, and so on); X-phrases can have subjects. To put it crudely, the X-bar theory of phrase structure could be innate. Second, since the meanings of parents’ sentences are usually guessable in context, the child could use the meanings to help set up the right phrase structure. Imagine that a parent says
The big dog ate ice cream
. If the child has previously learned the individual words
big, dog, ate
, and
ice cream
, he or she can guess their categories and grow the first twigs of a tree:

In turn, nouns and verbs must belong to noun phrases and verb phrases, so the child can posit one for each of these words. And if there is a big dog around, the child can guess that
the
and
big
modify
dog
, and connect them properly inside the noun phrase:

If the child knows that the dog just ate ice cream, he or she can also guess that
ice cream
and
dog
are role-players for the verb
eat. Dog
is a special kind of role-player, because it is the causal agent of the action and the topic of the sentence; hence it is likely to be the subject of the sentence and therefore attaches to the “S.” A tree for the sentence has been completed:

The rules and dictionary entries can be peeled off the tree: