SCIENCE: SURVIVAL OF THE CHATTIEST

For centuries the origin of language has divided scientists. Now a new Darwinian theory is being proposed. But how can this make sense when our ability to talk depends on co-operation, and not competition?

SEVEN thousand tongues are spoken today, it's said, and half a million may have come and gone since humans acquired the faculty of language, according to the Oxford biologist Mark Pagel. In their attempts to work out how that transformation might have occurred, scholars seem to have deployed comparable numbers of theories, perspectives, papers and bits of jargon. There are noun phrases, generative grammars, voice onset times and fricatives. There's the question of the descent of the larynx, the heated debate over the Neanderthal hyoid bone, and the bitter controversy over the australopithecine lunate sulcus. But, according to anthropologists Chris Knight and Camilla Power, the questions that mattered most to our distant ancestors, as they hesitantly entered into the domain of language, were much the same as those which matter most to us: "Do you really mean it?" and "Can I rely on you?" First and foremost, language is a matter of trust.

Knight, at the University of East London, and Power, at University College London, are developing a model which sets human language in a modern Darwinian framework. In the process, they are drawing together some of the most dynamic lines of argument in current British evolutionary thought. They will be presenting their latest ideas in London next week at the Second International Conference on the Evolution of Language. Taking place at City University, the event has been organised by James Hurford, of Edinburgh University, Jean Aitchison, the Oxford professor who gave the 1996 Reith Lectures on "The Language Web", and Chris Knight.

From a Darwinian point of view, language is tricky to explain because it brings up the problem of altruism. Language depends on co-operation, but natural selection is a matter of competition. Modern Darwinism is based on the principle that evolution doesn't take place for the benefit of the group. Natural selection acts on individuals, not groups. If a trait is good for an individual - that is, it helps the individual get more of its genes into the next generation - it will be selected, even though it may be bad for the group as a whole. Yet everywhere one looks, one sees co-operation flourishing in co-existence with competition. Modern Darwinism has devoted enormous effort to working out how self-interest can encourage one individual to help another. Its theories revolve around two concepts. One is inclusive fitness, the extent to which individuals share genes and therefore genetic interests. The other is reciprocal altruism, the technical term for exchanges in which A scratches B's back and B scratches A's.

The trouble with such arrangements is that they are vulnerable to cheating. If B suddenly refuses to scratch A's back in return, the sensible thing for A to do is to refuse to scratch B's back any more. Not only has the cheat come out ahead, but the system has broken down. The relative advantages of co-operation and cheating have been explored in extensive mathematical modelling exercises: these have shown that increasing the benefits of co-operation is not an answer, because this just increases the potential pay-off for cheating.

Backscratching, in the literal rather than the metaphorical sense, has proved to be a stable system of behaviour among primates, who devote a large amount of their time to mutual grooming. Primates that groom each other are primates who are well disposed to each other. Grooming can therefore communicate valuable information about the relationships between individuals in a group - who's grooming whom - and can serve to maintain social bonds within it. According to Robin Dunbar, of the University of Liverpool, language is, in fact, a superior form of backscratching. As groups increase in numbers, however, the time needed to carry out all the necessary grooming becomes impracticably large. In humans, Dunbar argues, language evolved as a means of "grooming" more than one individual at a time. The information it carried was similar in theme to that conveyed by grooming: who's friends with whom, who's fallen out with whom, who's doing what with whom. In other words, the prime purpose of talking is gossip.

Dunbar's model is part of a broader tendency to emphasise the role of social factors in the evolution of human mental capacities. Traditionally, human intelligence was assumed to have been driven by the relationship between humans and their artefacts, mainly those assumed to have been used to kill animals for food. We became what we are, it was felt, largely because of men and their tools. Now, women and relationships have moved to centre stage. Approaches to the Evolution of Language, the book of the first Evolution of Language conference (edited by James Hurford, Michael Studdert- Kennedy and Chris Knight; to be published later this year by Cambridge University Press) observes that scientists have historically tended to see language through the lenses of societies preoccupied by technology, as a means of communicating information about tools, hunting and similar practical matters. The new book sees itself as a product of the age of mass democracies, and human intelligence as essentially social. It treats language as a means of communicating about desires and motives, forming alliances and making friends; with successful feeding and mating as the end result.

In sum, it's what life is all about. The perspective is not trivialised by pointing out that it is also the product of the age of mass market soap opera, whose concerns are identical, apart from the feeding.

The attraction of gossip as an alternative to grooming is that it reaches more individuals at lower cost. But therein lies its great weakness, as Camilla Power points out. Grooming works as a system of exchange because choosing to groom an individual indicates both preference and commitment. The more individuals you can "groom" at a time, the lower your commitment is to any one of them. The cheaper the signal, the less reliable it is, and vice versa.

This axiom has been put to bold use by Amotz and Avishag Zahavi, radical Darwinian thinkers who have derived what they call the Handicap Principle from their field studies of birds in Israel, and since the mid-1970s have applied it to almost every conceivable form of communication. A peacock's tail is an "honest indicator" of fitness and genetic quality, because a bird has to be fit in order to be able to thrive while carrying such a massive burden. Similarly, primate grooming is a reliable signal of friendliness precisely because it takes time and effort. The trouble with words is that they are cheap.

Social intelligence is often called "Machiavellian intelligence", which was the title of an influential book published 10 years ago. Its editors, Richard Byrne and Andrew Whiten, noted in last year's sequel (known to them as Mach II) that they mean more by Machiavellianism than simply deception and exploitation. Rather, it refers to "relatively complex strategies fulfilling [ultimate] personal gain". These strategies can easily include co-operation, or behaviour from which other individuals gain greater benefits than costs. Whiten and Byrne draw on the authority of The Prince, in which Niccolo Machiavelli states that it "is useful, for example, to appear merciful, trustworthy, humane, blameless, religious - and to be so - yet to be in such measure prepared in mind that if you need to be not so, you can and do change to the contrary".

For a prince among men, language is the perfect Machiavellian instrument. It is cheap and infinitely flexible. It is also highly efficient, because it is established on a basis of trust. The solidity of this foundation is attested by the amount of deception and cheating that the system can contain without collapse. Chris Knight suggests that even if a group of particularly clever chimpanzees were to work out a set of signals equivalent to words, they would fail to establish a stable language system. The opportunities for deception would be too great. Every time one of these superchimps said "There's a leopard over the hill" - or "I'll give you some meat if you have sex with me" - the reaction would be "Why should I believe you?" For any statement which could not be immediately verified, "you have my word for it" would be the only possible response. And words, being so inexpensive to the speaker, would be next to worthless on their own. Since the cost of paying attention could so easily be higher than advertised, nobody would listen to what anybody else had to say.

Within the limits of an immediate frame of reference, however, a rudimentary symbolic signalling system could arise. Merlin Donald, a psychologist, has suggested that in the earlier stages of human evolution, vocal sounds and gestures would have been combined in a system which depended heavily on mime. These would have been very costly compared to words. As anybody who has played charades will recall, it takes an awful lot more effort to convey the message by mime than it does by speech. In a system like this, communication would be held back by the need to maintain high cost as an indicator of reliability. Where the truthfulness of such messages could be checked on the spot - "look, there's a leopard" - such a system could work. But there would be no way of communicating reliably about things that were not visible. That would have excluded objects which existed but were out of sight, things which no longer existed or had yet to come into existence, abstractions, or entities which existed only in the imagination. No past, no future, no dead people, and no gods.

Several religious traditions link language to divinity. In the beginning was the Word, says the Bible; the Indian deity Indra is said to have created articulate speech; similar themes occur in Norse mythology, while Plato has Socrates saying that the gods gave things their proper names. Chris Knight also maintains that the relationship between language and the divine is fundamental. In order for a stable, fully developed system of language to arise, he argues, the cheap signals of words needed to be backed by a set of costly signals. These, he believes, were provided by ritual. Language and ritual evolved together.

Here Knight is building upon the theory he has been developing for many years, an extraordinary vision of the origins of culture in the symbolism of meat, moonlight and menstrual blood. Based on the idea that females exerted collective, ritualised pressure on males to provide them with the meat needed to raise their young, it combines the hallucinatory imagination of a shaman with the cost-benefit rigour of a Treasury economist, and unsurprisingly the world was not really ready for it when it was published as Blood Relations in 1991. Seven years later, the advance of neo-Darwinian theory has made it seem a lot less outlandish as a means of examining problems in human evolution, if not as an account of actual events. With the new work on language, Knight and his colleagues are at last threatening to break into the mainstream. There are intriguing connections to be made with other strands of evolutionary thought. At next week's conference, for example, a paper to be given by Derek Bickerton of the University of Hawaii will propose that language depends on a mental apparatus which evolved to detect cheating, a faculty which the prominent evolutionary psychologists Leda Cosmides and John Tooby have suggested to be a significant influence on our thought processes.

Rituals are worthless if they are not difficult and costly. In our society, the principle is better appreciated by Muslims, obliged to pray five times a day, or Orthodox Jews, with their complicated and demanding observances, than by the remaining Anglicans, glancing at their watches if the sermon lasts more than five minutes. Many religions insist upon initiation rites, which may be extremely painful. Subincision, for example, practised by certain Australian Aboriginal groups, entails slitting the underside of the penis up to the urethra. In the light of signal theory - not to mention the Handicap Principle - this extreme form of ritual mutilation makes sense.

The group imposes ritual on its members to verify their commitment, and to bind them in bonds of mutual trust. With trust guaranteed, it becomes possible for the members of the group to communicate using cheap signals. It also becomes possible to create symbols which do not correspond to things in the physical world - such as gods. The ritual is fundamentally an enactment of moral authority. In making language possible, this moral authority reappears as the Word, and, of course, the Word is God. Knight compares words to banknotes, which "promise to pay the bearer on demand". Ritual guarantees the value of words in the way that banks, backed by the law and other institutions, guarantee the promises made by banknotes, which would otherwise just be pieces of paper. As it says on every dollar bill, "In God We Trust".

! The Second International Conference on the Evolution of Language takes place at the Tate Building, City University, Goswell Road, London EC1. The event runs from 12.30 to 6pm on Monday 6 April, 8.30am to 5pm on Tuesday and Wednesday and ends at 1pm on Thursday 9 April. Admission is pounds 70 for the whole event; pounds 15 for Monday or Thursday alone; and pounds 20 for Tuesday or Wednesday alone

HAVING THE LAST WORD

IF YOU want to bait a bunch of linguists, ask them about the origins of language. They have a long-standing aversion to the issue, dating back to the days when imaginative thinkers would suggest, for example, that humans learned speech from birds. By 1866, they were so exasperated that the Linguistic Society of Paris banned papers on the subject.

Individual languages, as distinct from the language faculty itself, were another matter. In 1786, Sir William Jones realised that there were similarities between Sanskrit, Greek and Latin, and that they must have a common ancestor. His insight gave rise to the great 19th-century enterprise of comparative linguistics, in which scholars (including Jacob Grimm, better known for the fairy-tales he collected with his brother) traced relationships between languages by mapping similarities between them. They succeeded in reconstructing something of the putative ancestor of Sanskrit and the European languages, known as Proto-Indo-European. More recent efforts have attempted to recover elements of even older languages, such as Nostratic, said to have been spoken as much as 15,000 years ago, and even Proto-World, supposedly the mother of all tongues. Because languages change so quickly, however, many linguists doubt whether reconstructions can reach back more than about 10,000 years.

In 1957, the cognitive sciences were revolutionised by Noam Chomsky, who identified deep structures common to all languages, amounting to a "Universal Grammar". He argued that children acquire language so easily that they must have neural machinery specialised for the purpose. In doing so, he laid the foundation for the discipline of evolutionary psychology which has emerged over the past decade, and considers that our minds are made up of a whole collection of such specialised "modules", or organs. Yet Chomsky himself has been averse to the idea that evolution produced language. Following his line, the palaeontologist Stephen Jay Gould has suggested that language was an accidental by-product of the human brain's large size.

The campaign to have language recognised as a product of natural selection began in earnest with a 1990 paper by Steven Pinker and Paul Bloom; Pinker's book The Language Instinct sets the arguments out for lay readers. An alternative evolutionary position is that of neuroscientist Terrence Deacon, whose book The Symbolic Species agrees that language is the result of natural selection, but denies that it needs a special organ or instinct.