The naming insight is followed by a 'naming explosion'. Names come popping out of children like stars out of fireworks. This eruption in vocabulary leads to word combinations, mummy push, daddy car and so on. Some phrases are novel, as byebye sock, allgone kitty, which are unlikely to have been copied from adults. Recurring patterns are found as with sand toe, 'I've got sand in my toes', sand eye 'I've got sand in my eye', and sand hair 'I've got sand in my hair'. The parents were probably too busy mopping the sand off this child to admire the consistency of its language rules.

Jean Aitchison

Figure 2. Innately guided behaviour.

Youngsters look for regularities in language, as shown by the wug-test, devised by Jean Berko Gleason in the 1950s. 'Here is a wug', she said, showing a picture of a bird-like creature. Then she showed two of them: 'Now there are two...' 'Wugs' responded children from a very young age.

Children do not always get it right first time. Two-year-old Sophie learned the words broken, fallen and taken. She wrongly concluded that English past tenses end in -en. She then invented a whole range of new past tense forms, such as boughten, builden, riden, gotten, cutten, wanten, touchen, maden, tippen, as in 'me tippen that over'. Sophie gradually dropped these -en forms--probably when she discovered the normal past tense for each verb. Children dislike finding two words which mean exactly the same thing, and usually drop one of them.

By the age of three, children utter long sentences, though some things, such as pronouns, still cause problems. Three-year-old Adam said his doll 'shuts she's eyes', instead of 'shuts her eyes'.

At around three and a half, children talk freely. By this time, they have acquired most of the constructions used by adults. This is true of monolingual children, and also bilingual ones. A few gaps still exist for all children up to the age of around ten, and word-learning goes on throughout life.

This predictable sequence of events is typical of biologically scheduled behaviour, as pointed out by Eric Lenneberg, a pioneer in this field. His book Biological foundations of language, published in 1967, was a major landmark. Before then, natural behaviour, such as seals swimming, was usually separated from nurtured or learned behaviour, as when seals can be taught to jump through hoops.

Lenneberg showed that this divide is over simple. Most natural behaviour requires some learning: pigeons naturally fly, but they have to spend time learning how to stay in the air. Conversely, learning would be impossible if it did not build on natural talents: pigeons can be trained to distinguish between letters of the alphabet, but only because they already have acute eyesight.

Language is an example of maturationally controlled behaviour, Lenneberg pointed out, behaviour which is preprogrammed to emerge at a particular stage in an individual's life, provided the surrounding environment is normal. Walking and sexual behaviour are other examples. Such behaviour emerges before it is critically needed, yet cannot be forced to appear before it is scheduled. Some learning is required, but the learning cannot be significantly speeded up by coaching. No external event or conscious decision causes it, and a regular sequence of milestones can be charted.

An ability to cope with language structure is largely separate from general intelligence. In recent years, several so-called 'cocktail party chatterers' have been discovered--children who have a non-verbal IQ so low that they may not even know their own age, but who speak fluently. As at cocktail parties, they talk for the sake of talking, and their speech may not make sense. Take Laura, an American teenager: 'I was sixteen last year, and now I'm nineteen this year' or 'It was no regular school, it was just good old no buses.' Such chatterbox children are not just repeating set phrases, because they make grammatical mistakes which they are unlikely to have heard, as in Laura's statement that 'Three tickets were gave out by a police last year.'

Just as bees learn fast to distinguish flowers from, say, balloons or bus-stops, so human children are preset by nature to pick out natural language sounds: they do not get distracted by barking dogs or quacking ducks. Their learning is innately guided (see Figure 2). Inbuilt signposts direct youngsters, so they instinctively pay attention to certain linguistic features, such as stressed vowels and word order. Children's main task is to discover which of these features have priority in the language or languages they are acquiring, just as bees have to learn whether to look for heather, roses or lilies.