TRENDS in Genetics Vol.18 No.2 February 2002 http://tig.trends.com 0168-9525/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0168-9525(01)02596-3 57 Research Update Research News A point mutation has recently been found in a gene from affected members of a family w ith an autosomal dominant pattern of inheritance for specific speech and language impairment. However, this does not mean w e have localized the ‘gene for language’. The phenotype is complex, and the affected gene, which is concerned w ith regulating activity of other genes, is common to human and mouse. The discovery is nevertheless important, because it w ill help us to identify target genes that play a role in development of the neural circuitry involved in language. One of the most surprising discoveries of r ecent year s has been t he degr ee t o whi ch genes ar e conser ved acr oss speci es. I t i s sober i ng t o di scover t hat t her e i s mor e than 98% overlap between the human genome and that of our nearest evolutionary neighbour, the chimpanzee. The puzzle is to work out how such a small genetic difference can account for such vast differences, especially in brain development and cognitive function [1]. Language is the most striking instance of a complex, human characteristic that has no equivalent in other primates. Back in the 1950s, the linguist N oam Chomsky drew attention to some remarkable features of the human language capacity, concluding that the ability of children to learn language had to be explained in biological terms, rather than just as a product of social learning [2] (Box 1). Mathematical learning theory has been used to argue that children could not learn grammatical rules simply from exposure to spoken language: they must therefore come to the language- learning task with brains that already have some implicit knowledge of language structure. This knowledge would have to be highly abstract to account for children’s ability to learn any l anguage t hey ar e exposed t o, i n spi t e of substantial superficial differences in the grammar. Comparative linguists aim to define the common core principles that are shared by all human languages (i.e. a ‘universal grammar ’). This, according to Chomsky, must be innately specified in the human brain. I n hi s book The Language Instinct [3], Steven Pinker developed Chomksy’s arguments along evolutionary lines, drawing an analogy between human language and the trunk of the elephant. The trunk has clearly evolved from an ancestral nose, but its functions go well beyond what we expect noses t o do. ‘Language,’writes Pinker, ‘is obviously as different from other animals’ communication systems as the elephant’s trunk is different from other animals’ nostrils’(p. 334). Just as the trunk is a new organ, forged out of old parts, so language will depend on modifications to brain circuits that might have served other functions in our ancestors. However, according to Pinker, the Putting language genes in perspective Dorothy V.M. Bishop The puzzle of language acquisition is how children rapidly learn the grammatical structure of their native language without overt instruction or, indeed, without explicit awareness of the underlying rules [a]. Chomsky noted that a simple associative mechanism that detected statistical regularities between word sequences would be unable to extract the recursive structure of language. Consider, for instance, the sentence ‘the pencil on the shoe is grey’. If we simply related each word to the next, we would conclude that the shoe was grey. To understand this sentence we must recover the underlying structure, in which the phrase ‘on the shoe’ is embedded in the main proposition, ‘the pencil is grey’. Chomsky further argued that word meanings cannot be used to deduce grammatical rules, because we are able to distinguish grammatical from ungrammatical sequences, even when these are completely meaningless, as in: • Colourless green ideas sleep furiously (grammatical) • Sleep green colourless furiously ideas (ungrammatical) Furthermore, a sentence can be ungrammatical but meaningful, as in: • The boys play football yesterday Chomsky marshalled two additional arguments from biology to make his case that language acquisition is special, and not explicable in terms of general learning mechanisms. First, attempts to teach non-human primates grammatical language have been singularly unsuccessful. Second, language is mediated by specialized brain regions in the left hemisphere. Although such arguments for an innate language faculty have been influential, they have not gone unchallenged [b]. Psychologists have questioned whether young children really do know abstract grammatical rules. It has been proposed that language acquisition initially proceeds as a process of piecemeal rote learning of associations between word strings and meanings, with implicit awareness of more abstract grammatical regularities becoming apparent only at a relatively late stage in the learning process. Although nobody has yet devised a neural network that can learn a human grammar, associative learning mechanisms with a parallel distributed architecture can do far more than simply learn sequential associations. Furthermore, regional brain specialization for cognitive abilities can develop as skills are acquired, and does not necessarily indicate innate prespecification of brain areas for specific functions. Indeed, such specialization is seen for skills such as reading, which is a relatively late arrival on the human cultural scene, and could not plausibly be seen as a hard-wired brain function arising through natural selection. References a Chomsky, N. (1988) Language and Problems of Knowledge: The Managua L ectur es, M I T pr ess b Elman, J. et al. (1996) Rethinking Innateness. A Connecti oni st Per specti ve on Devel opment , M I T Pr ess Box 1. Is language special?