In acquiring language, children acquire thousands of words. What they are doing is establishing thousands of connections between phonological, semantic and syntactic information.

The first evidence of this process emerges around 9 to 12 months when children make a consistent connection between a form they hear and a thing or action to which it refers. They may point to the appropriate creature on hearing the word ‘tiger’, or move their hands appropriately on hearing the word ‘clap’. By the time they understand 50-150 words, around 1 to 1½ years, they start producing words (see Ingram 1989). Their lexical acquisition then proceeds apace. It is estimated that they know some 8,000 root words by the age of 6, which works out at an average rate of five new words a day, assuming they kick off at 18 months (see Carey 1978 and Clark 1993).

Children absorb words like sponges. How are they doing this? They are not presented with vocabulary lists which pair phonological forms and meanings. They start from the scenes they observe and participate in, and the stream of speech they hear. In order to acquire words, they must segment that stream of speech, separating out the sound pattern of each word. They must segment the scenes in which the speech occurs, separating out the aspects of the scene which are picked out by the words. And they must map the one onto the other, making connections between the sound patterns they have segmented and the aspects of scenes they have segmented.

Construed in this way, the child's task looks daunting. Imagine a child hearing the utterance ‘There's a tiger’ as he looks at a striped creature strutting round (see illustration).