BOOK REVIEW / The tip of our tongues: The language instinct by Steven Pinker, Allen Lane/Penguin Press pounds 20

'IS thought dependent on words? Do people literally think in English, Cherokee, Kivunjo, or, by 2050, Newspeak? Or are our thoughts couched in some silent medium of the brain - a language of thought, or 'mentalese' - and merely clothed in words whenever we need to communicate them to a listener? No question could be more central to understanding the language instinct.'

This is Steven Pinker, an experimental psychologist at the Massachusetts Institute of Technology. His answer is unequivocal: 'mentalese'. Language, he says, does not shape thought: thought shapes language. Fundamentally, human minds work in the same way, regardless of the language we happen to speak; and this working is innate - we cannot help but learn to talk. To a visiting Martian, Earthlings would appear to speak one language, though with mutually unintelligible vocabularies. Language universals, ie a universal grammar at the deepest mental levels, predominate over language differences, which are surface structures.

It is of course the approach of MIT's most famous thinker, Noam Chomsky, who revolutionised the study of language in the late 1950s. Chomsky is the greatest influence on Pinker; indeed, he is among the 10 most cited writers in all the humanities. But how many people really understand Chomsky? His writings perfectly fit Mark Twain's definition of a classic, Pinker notes: a book which everyone wants to have read but nobody wants to read.

The Language Instinct sets out to explain Chomsky, along with some other ideas less acceptable to him. It is a challenging, sometimes rambunctious read, made viable for the non- specialist by its entertaining style and cullings from ordinary speech, children's early speech, ambiguous newspaper headlines, popular songs and literary works. (Pinker analyses, for instance, why we cannot settle on a plural for 'Walkman': is it 'Walkmans' or 'Walkmen' or what?) The book's firm basis in three or four decades of laboratory studies - of children and adults, speech impairments and brain function - gives it a concreteness and accessibility lacking in the baffling abstractions of Chomsky's 'syntactic structures'. Anyone wanting to grasp the basics of Chomsky and the present state of cognitive science should read Pinker. But his own thesis did not convince me.

Pinker stoutly maintains that the ability to speak is no different in origin from, say, the ability to see. The organs of speech and the eye have evolved, he says, by natural selection, as first envisaged by Darwin. Language is unique to humans, he concedes, unlike eyes, but that does not bother him: 'A language instinct unique to modern humans poses no more of a paradox than a trunk unique to modern elephants.' Further on, he claims that the language instinct impresses him no more or less than the island-building instinct of corals or the photosynthesizing instinct of bacteria. Yet, almost in the same breath, he admits: 'Most objects in the universe - lakes, rocks, trees, worms, cows, cars - cannot talk . . . How are we to account for this miracle?'

Without new theories, is Pinker's reply to his own question. In a nutshell, he argues that language and language organs evolved because those of our ancestors who could communicate vocally were more likely to survive than those who could not. He knows there is no real evidence, archaeological or otherwise, and he cannot propose a theoretical mechanism for the momentous change; what's more, he is certain that language evolved only once. It won't do - and Chomsky, as befits a wiser mind, does not accept it. In 1988, he wrote: 'Evolutionary theory is informative about many things, but it has little to say, as of now, about questions of this nature.' The latest paper by Chomsky cited by Pinker is entitled, aptly enough: 'Linguistics and cognitive science: problems and mysteries'.

Mystery - it is a word seldom used by Pinker, like 'consciousness'. By divorcing language from thought so emphatically, he leaves himself free to treat language as something much simpler than we instinctively feel it to be. Though he frankly admits the failings of computers to decode the meaning of simple sentences, he remains keen to see the computer as a model for the human brain. Repeatedly he uses mechanistic terms for the functioning of language in the brain, such as 'engine', 'wiring', 'machinery', 'circuitry', 'organ of computation', and he calls grammar 'a form of mental software'. He hopes that research will eventually identify specific parts of the brain responsible for specific aspects of language - even, perhaps, a 'grammar gene'.

As for reading and writing - mysteries of visible language fully on a par with speaking - they get very short shrift. No one has much idea of what is going on in our minds when we read and write, especially when we write - but we may be sure that whatever it is, vision, language and thought are intimately entwined. Pinker's approach to writing is both vestigial and parochial. His remarks on writing systems, especially Chinese, are seriously misleading, and his bibliography contains not one academic work on writing. It is ironic that linguists, who rely on the printed page to communicate their ideas, can still dismiss writing as 'an artificial contraption connecting vision and language'.

My plea is not 'mystery for mystery's sake'. I applaud Pinker's scientific urge to get to the bottom of language. The evidence of its complexity that he presents - such as anomic patients who cannot recall certain classes of noun, eg fruits and vegetables, but otherwise speak normally - is invaluable and fascinating. But his theories, like so much evolutionary theory, beg even larger questions than those they are designed to answer.