Me Jane, you verbal lizard?

Philip Lieberman's message is that "the theory that Charles Darwin proposed can account for the evolution of [human language]". Language is not biologically special or based on some single biologically recent development that defines a universal grammar and differentiates man from other species. Lieberman marshals neurophysiological, palaeontological and other evidence to argue that language exploits neural mechanisms that evolved to serve other functions (for example manual motor control), and some of which reside in the most ancient structures of the brain, the subcortical basal ganglia or the "reptilian brain".

Lieberman urges that language must be predominantly a learnt rather than innate ability. Such abilities as walking upright, which is shared by all adult humans, might be supposed to be uncontroversially innate. But, according to Lieberman, this should also be seen as a learnt skill.

All this contradicts the assumptions of the Chomskyan brand of linguistics. Lieberman criticises not only Noam Chomsky but his followers Derek Bickerton and Steven Pinker. Their claims that language behaviour is governed by special-purpose modules of brain structure are false, according to research cited by Lieberman, and owe more to misleading analogies with computer science than to empirical evidence. Lieberman sees Chomskyan linguistics as a "worldwide religion", rather than a scientific enterprise; it protects its standing by isolating itself from neighbouring disciplines that might come up with counter-evidence.

The irony is that Lieberman used to be a leading advocate of such unscientific ideology. In a 1975 book, he argued that the Levallois tool-making culture showed that early man had acquired the neural basis of transformational grammar, because the two-stage Levallois "core and flake" flint-working technique resembled the two-stage process of creating a deep sentence structure and transforming it into a surface structure. Lieberman now says: "The analogy, in retrospect, was forced" - surely an understatement.

Today, the trickle of apostates from the Chomskyan church is turning into a stampede, as people who managed to live with the logical flaws of nativist linguistics realise that techniques such as corpus research and Pet and MRI scanning are delivering hard data about language usage and neural processing that render redundant, and often contradict, the purely speculative premises on which that style of linguistics rested. Thus Lieberman wants to get his recantation on record before he occupies the undignified position of a priest without a congregation.

One consequence is that long passages are devoted to unconvincing defences of the line taken by Lieberman in earlier writings. He argued, from fossil evidence, that Neanderthalers lacked human-type language. He now describes this as a misunderstanding, saying that he made the more limited claim that Neanderthalers could not produce the full human range of vowel sounds. Yet he still wants to treat this as explaining the Neanderthal extinction. Lieberman cites findings that gene-pool boundaries tend to correlate with boundaries between languages. He argues that a distinctive Neanderthal accent would have made them an isolated population, vulnerable to extinction. But the explanation for the gene pool-language correlation is known to any young man: girls expect some talking. Contrary to cartoons depicting cavemen dragging women off by the hair, it seems that in the Stone Age as now, a man who could not communicate verbally with a woman was unlikely to mate with her. Lieberman's claim about Neanderthal vowels does not require dialect differences between Neanderthalers and anatomically modern man to have been much greater than those between northern and southern England nowadays, which are no bar to amorous relationships.

Other passages offer an idiosyncratic view of the history of phonetic science. Is it really true that 19th-century research on laws of sound change related to choice between articulatory and acoustic descriptions? Lieberman apparently believes that before Alexander Melville Bell's Visible Speech of 1867, nobody knew how to classify consonants by place of articulation - he wrongly attributes that book to Alexander Melville's son Alexander Graham, inventor of the telephone. (The book has a number of errors that a publisher's editor could not be expected to catch, relating for instance to proper names.)

Despite its shortcomings, the book draws together references to recent biological research having implications for linguistics. Lieberman rightly urges that linguists ought to be paying more attention to this work.

Geoffrey Sampson is professor of natural language computing, University of Sussex.