What is the best way to teach a baby to talk? Nobody really knows, but most scientists agree that it probably doesn’t matter. Unless they are completely isolated from human speech, children learn to speak pretty well. However scientists from different traditions differ widely in their explanations for why this is so.
One view, widely held by linguists and expounded with great clarity by Steven Pinker of Massachusetts Institute of Technology in an extremely readable book*, is that the apparatus for language is embedded in the human brain and develops spontaneously. All that is necessary for the baby to learn to speak is to fine-tune the language machine inside his or her head to idiosyncrasies of the local dialect, and to learn the vocabulary. The normal development of the brain and body does the rest.
Evidence supporting this view comes from linguistics, neuroscience, genetics and psychology. However, it does not rule out a completely contrary view, held by cognitive psychologists. This is that a child works out how to speak by understanding events that surround it and imitating the language that describes them.
One of the strongest arguments advanced by the linguists is that, in every language, the rules for generating meaningful sentences correspond to minor variations of a universal grammar. The grammar has nothing to do with questions like whether we split infinitives or end sentences with prepositions. It simply specifies how words can be combined to produce phrases and how phrases are combined to produce sentences.
There are no universal rules about how to make words. Except for onomatopoeic words like “bow wow” there is no way of deducing the meaning of a word from its sound. The same word can have several unrelated meanings and new words can be (and are) invented whenever we need one. However, if I invent a new word it means nothing to you. On the other hand, we are both able to generate completely new sentences and to understand them instantly, because we follow similar rules for their construction and interpretation.
The number of sentences in any language is limitless. Any of us could generate many times more sentences that we would have time to speak during the course of our life. Whenever we learn a new word we can immediately deploy it correctly in new sentences. Linguists argue that we could not generate novel sentences simply by repeating what we hear, just as we cannot converse in a foreign language merely by using a phrase book. We need a machine inside our head that knows the rules for generating – and decoding – completely new combinations of words.
If language depends on such a machine, it should be possible to track it down in the brain. It is. A stroke or any other injury that damages a small area, usually on the left-hand side of the brain, leaves the sufferer unable to speak or to write. Fortunately the organisation of the brain is flexible enough that, if the damage happens early in life the corresponding area on the other side of the brain takes over and the ability is preserved. This flexibility could indicate that the language machine is not unique, or simply that the genetic rules for building it can be applied to different parts of the brain.
There is evidence that genes affect the ability to learn language. Dorothy Bishop, of the Medical Research Council’s Applied Psychology Unit in Cambridge studies children who are slow to learn language but who have otherwise normal intelligence. Language difficulty runs in families. If one of a pair of identical twins has it there is about a 90% chance that the other will have it too. For non-identical twins the chance is about 45%. Although this is strong evidence for a genetic defect, Bishop thinks that the root of the problem may be in the ability to receive the language input, rather than in the language machine.
Bishop is also sceptical about the importance of the universal grammar . “I think the linguists are all asking the wrong questions. Children don’t learn the universal grammar, they learn something about how likely it is that one word follows another.” The apparently universal nature of the rules of language may simply reflect the limitations imposed by human memory and attention spans, she says.
One way to resolve this question, according to Julian Pine of Nottingham University would be to work out what sort of grammar children use from the way they say things. Is their speech governed by rules that they have learned from their observation of the world, or do they use an immature version of the universal grammar that is a part of their genetic inheritance? The available data can be interpreted either way.
Pine is just starting a project to collect more data. Fourteen children who are just starting to speak will be recorded for two hours every three weeks for a year. When the conversations are transcribed and analysed it may be possible to get an answer, but it won’t come quickly; the detailed analyses required are immensely time consuming. “Those data could keep me busy until I retire” says Pine, who is still in his early thirties. *The Language Instinct, Steven Pinker, Penguin 1994