Robin was sitting in the classroom, giving me an exact account of the Russian Revolution. It was his specialist subject and he knew every date, every manoeuvre, the names of the Bolshevik leaders, and where they were when the Winter Palace was stormed. Robin was 16.

“So, what do you think of Lenin?” I asked. He looked at me blankly. “I don't think anything of Lenin,” he said. Robin was autistic.

He had an extraordinary grasp of facts, meticulously arranged in his mind. He had no concept of analysis or interpretation. The idea of forming an opinion was alien to him. With that incapacity came social isolation, an inability to form friendships or any lasting relationship. He was stranded, with his brilliant but disabled mind. Bringing him up had been a constant strain for his parents.

Quite how he would fare in the wider world was not yet clear.

Robin, and thousands like him, are at the centre of an ethical debate with far-reaching consequences. Within a few years it may become possible for expectant mothers to have prenatal tests to determine if their child is likely to be autistic.

These may be genetic, to see whether the characteristics of autism have been inherited, or tests of amniotic fluid in the womb to detect high levels of testosterone that have been found to be associated with the condition - mainly in boys.

As the parents of the first British baby screened to be free of a breast cancer gene celebrate the birth of a healthy daughter, this must seem yet another miraculous step in the advance of science. For any family that has experienced the anguish of living with an autistic child, the prospect of being able to determine if another is about to be born would be invaluable. It would offer that most precious commodity - a choice.

Just as with Down's syndrome, cystic fibrosis or spina bifida, a mother-to-be could decide whether she can cope with the strain of bringing up a disabled child. With an autistic child, it may mean a lifetime of rejection - living with someone unlikely ever to fit into the family, who responds with blank incomprehension to affection, whose behaviour may be erratic and disturbing, whose condition is permanent. Autism, and its associated condition, Asperger's syndrome, can range from virtual incapacity at one end of the spectrum to the merely strange at the other. To bring up one autistic child is a challenge to the sanity of an entire family. To bring up two might destroy it.

The evidence of Down's syndrome suggests that very high numbers of mothers-to-be opt for an abortion if pre-natal tests show that their child has the condition. In America it is as a high as 90 per cent. In Britain, it is not so high and may be reversing - as knowledge grows, perhaps more mothers elect to keep their babies. There is, however, a critical difference between Down's and autism, highlighted by Professor Simon Baron-Cohen, director of the Autism Research Centre at Cambridge. He points out that autistic males often turn out to be skilled at mathematics and engineering - some reaching near-genius level. Almost all the mathematical giants of the past have been male. He says that Newton and Einstein were almost certainly autistic, finding relationships difficult. Artists, too, have suffered from autism or Asperger's - including the blind pianist Derek Paravicini, the artist Peter Howson and, reportedly, the film director Steven Spielberg. So if we found a test for autism, and gave parents the opportunity of aborting the foetus, we might eliminate not just an unwanted and difficult child but a potential genius.

Here lies the dilemma. Should medical science offer the opportunity to eliminate a child who may turn out to be, not only a valuable member of society, but an important contributor to its future? And here lies a further twist in the moral maze. If that were the decision, what would be the justification for deciding that only the most intelligent members of society should be protected, while the less able were judged expendable. Does not that come close to Nazi-style eugenics, the one aspect of genetic engineering we have all determined will never again be contemplated?

Professor Baron-Cohen says that we must debate these matters now, before even the possibility of a test becomes a reality. I have no doubt he is right. But I am far from clear which side we should be on.

Every human instinct must surely be against some form of national screening that would offer the opportunity to breed out the wild, the eccentric, the sometimes weird, crazed individualists who break free of routine constraints and offer the diversity on which we thrive. Can we afford to lose a future Einstein?

There is a deeper strain to the debate. Who is to judge where lies the dividing line between madness and norm? As Kamran Nazeer so brilliantly described in Prospect magazine last year, it is possible to convert the apparent drawbacks of autism into an ideal - to learn the art of conversation, for instance, and to become as adept at it as a “normal” member of society.

As the father of a bipolar son, whose understanding of his own condition and whose empathy with his fellow human beings far surpasses my own, I claim no superiority of intelligence when it comes to deciding who is rational and who not. So I shrink instinctively from any notion that we should be given the opportunity of discarding a future human being simply because he or she may be an inconvenience.

If that means holding back science or our knowledge of genetics, even at the expense of suffering families, I think it a price worth paying. To interfere with the natural diversity of the human race runs the risk of impeding natural selection itself. And that, in Darwin's bicentenary, would be a backward step.

Times Recommends

• Five golden rules for regulating banks
• Uighurs’ cry echoes round the world
• It is not just democracy that is illegal