How we learn to speak is a mystery. Neil Smith and Ianthi-Maria Tsimpli have studied a savant who speaks 20 languages but struggles to put on a shirt.

Readers of this newspaper probably consider themselves intelligent. Being good at history is no guarantee that one will shine at algebra, but it is a commonplace of the academic world that success in one area is a good predictor of success in others, and it is intelligence that is deemed to underlie success. Indeed, the rationale behind ascribing people an IQ is that intellectual and other abilities tend to cluster. That one may wish to distinguish verbal and non-verbal intelligence is not taken to detract from the value of the measure: people are still intelligent or not.

Christopher makes any such assumption problematic. He is unique. Aged 34, he lives in sheltered accommodation because he is unable to look after himself. He cannot find his way around, and has poor hand-eye coordination, so that everyday tasks like shaving or doing up his buttons are tedious chores. He fails to remember numbers (a task which most five-year-olds can do), and he has some of the characteristics of autism. Yet he can read, write, translate and communicate in 15 or 20 languages. Christopher is a "savant": someone with general dis- ability but "fragments of genius".

Is Christopher "intelligent"? On different IQ tests he scores anything from 40 to more than 120: from the severely retarded to good undergraduate standard. He is not proficient at noughts and crosses, but he can translate from Portuguese to Polish. On holiday in Mallorca he kept getting lost, but acted as an enthusiastic and much appreciated translator between German and Spanish for his fellow tourists. He cannot impute false beliefs to others (and does not lie himself), but he has a phenomenal vocabulary. Asking whether he is intelligent is not a sensible question. One thing Christopher makes clear is that intelligence is not a unitary notion.

É«ºÐÖ±²¥

His case casts light on other issues in linguistics and psychology. We have been working with him for six years and believe that he provides evidence for the nature of our knowledge of language, and for the structure of the human mind in general.

Although it was his prodigious polyglot talent that first brought Christopher to our attention, there is a striking mismatch between his ability in English and his proficiency in his other languages. His knowledge of English is just like that of other native speakers. Even though his own speech tends to the laconic, he has normal intuitions over the whole range of English sentence structure. He not only corrects simple mistakes of agreement and word order, but reacts appropriately to contrasts as subtle as that between "Which books did you throw away without looking at?" and the questionable "Which books did you throw away without looking at them?" When it comes to his 15 or so second languages, the situation is rather different. First, his fluency varies from language to language: his French, German, Greek, Hindi, Italian, Polish and Spanish, for instance, are much better than his Danish, Finnish, Norwegian, Russian, Swedish, Turkish and Welsh.

É«ºÐÖ±²¥

In syntax his abilities are less striking. Although he can translate and communicate with some facility, he makes many mistakes, seeming to filter all his languages through English. He can cope with differences of the kind seen in the contrast between English "she read the book" and Spanish "leyo el libro", where Spanish can, but English cannot, leave out the subject pronoun; but he consistently fails to judge correctly sentences like "Quien dijo que leyo el libro?" (literally: "Who did she say that read the book?"), presumably because the "that" in the English translation renders it ungrammatical.

Why should there be this double mismatch between his first and second languages on the one hand, and between the lexicon and the syntax on the other? Noam Chomsky's theory of language provides at least partial answers, and hence Christopher's case lends some support to Chomsky's position. For Chomsky, the acquisition of the syntax of our first language consists mainly in fixing the parameters of universal grammar. We are born knowing the full range of possibilities provided by the world's languages, and our task is to select the grammar for the one we are exposed to. Once the parameters are set, we can absorb new words, but our syntax is fixed. Christopher's English is normal: he acquir- ed his first language in the usual deterministic fashion but, despite his obsession with other languages, his learning of them is different.

That universal grammar is still operative, in his case as in that of everyone else, was shown when we taught him an invented language whose rules deviated from those made available by linguistic theory. We hypothesised that if we made up constructions which diverged from what is possible in human languages, Christopher should be unable to learn them, whereas undergraduate controls should be able to use their non-linguistic abilities to solve the problem by raw intelligence. Our predictions were generally borne out, with the added twist that, when we invented a construction in which the subjects had to count words, not even the students could cope.

Christopher shows us that Chomsky's theory of grammar is well supported; he also shows that assigning someone, anyone, a single IQ is silly.

É«ºÐÖ±²¥

Neil Smith and Ianthi-Maria Tsimpli, University College London and University of Cambridge. The 'Evolution of Language" conference will be held in Edinburgh University from April 1-4.