A fresh look at the role of language has led to new ideas about how it all started. Traditionally, humans are regarded as reliable fact-swappers, and conveying information is often claimed to be the primary purpose of language. This view was put forward by philosophers, such as the seventeenth century thinker, John Locke, who stated that language is 'the great conduit, whereby men convey their discoveries, reasonings, and knowledge, from one to another'.

But this idea is misleading. Language is good at transferring some types of data, especially negative reports, such as: 'No buses will run on Sunday', or 'The milk hasn't arrived' - provided that the speaker is telling the truth.

But it's bad at other types, especially spatial information, where instructions such as 'Take the third turning on the right then the fourth on the left' would be much clearer on a map. And as for tying knots, Hilaire Belloc once said: 'If you can describe clearly without a diagram the proper way of making this or that knot, then you are a master of the English tongue'. Perhaps he should have said: 'You will have considerable difficulty describing a knot, however many languages you have mastered'. Language is also bad at conveying pain or emotion. This patchwork of efficiency and inefficiency is fairly typical of behaviour that is biologically programmed: it has evolved to deal with some things, but not others, just as rabbits nibble grass but don't crack nuts.

Early human language was probably not a fact-swapping device. Its original role can be uncovered by looking at behaviour which we share with our ape relatives, according to anthropologists.

Jean Aitchison, The Reith lectures: The language web, 1996