Another refinement of Darwinism involves an increase in the flow of raw material upon which it feeds. To damage DNA can produce new genes ready for use by an alert technologist. Penicillin once depended on tiny amounts of antibiotic made in vast vats of fungus. Breeding from the most productive strains gave a hundredfold increase. The next step did much more: mutations caused by radiation and chemicals led to a new generation of antibiotics, never seen in nature.

An even better way to renew the fuel for selection is to import genes from other species. One of the successes was the new crop triticale, a hybrid between wheat and rye. It can grow in dry places and is of benefit to agriculture in places (such as the American Great Plains) low in rainfall. It demonstrates the gains to be made by even a modest investment in moving genes between species. Another approach is to turn to a domestic plant's untamed relatives, as has been done with wheat itself by crossing with wild grasses that contain genes of value on the farm.

The standard agricultural approach of breeding from the best — evolution writ large — has limits, which are soon reached. Many crops and farm animals can evolve no further because they have used up their genetic reserves and have no source from which to replenish them. The constraint is set by sex: by the fact that to make creatures with new mixtures of genes their parents must mate. In spite of occasional lapses in the plant world, there are strict biological controls as to who mates with whom. The partners must be of different sexes but the same species. A few modest exceptions — triticale being one — are allowed: but to recombine genes, in nature or on the farm, sex is unavoidable. That law much decreased the ambitions of evolutionary engineers because genes that might be useful in improving one form are locked away within another.

Agriculture itself began with some mild infringements of sexual convention. Farmers ameliorated nature by clearing trees to allow vegetation to flourish. Plants that never normally meet came together and, from time to time, hybrids appeared. They contained combinations of genes never seen before. The process goes on. Many mudflats around Britain are covered by a tough grass, a hybrid between a local species and one introduced from America. The new mixture of genes does better in a harsh environment than does either parent, and has become a pest.

Chromosomes show that modern wheat began when two species of grass (each of which is still used for food in the Middle East) hybridized. As on the mudflats, the new cross was more productive than either parent. Soon, another grass crossed with the new recombinant, improving it further. This was the predecessor of every one of the billions of wheat plants grown today. The early farmers had moved chromosomes, genes and DNA from one species to another. They were the first genetic engineers.

Now, science has made sex universal. Molecular biology allows genes to be shifted among lineages which were once quite alien to one another; to make recombinant DNA not by the joint efforts of male and female, but by bypassing the inconvenience of reproduction altogether. Genes can be moved from more or less anywhere to anywhere else. At last, DNA can be used where it is needed, wherever it comes from. The biological rules have been broken and a new era of agriculture is at hand.

Genetic engineering began in bacteria, which have a commendable range of sexual interests. They exchange information in many ways; by taking up naked DNA, by a process of mating rather like that of higher animals and by the use of a range of third parties or viruses. This 'infectious heredity' (which suggests that venereal disease evolved before sex) has been subverted by science. The gene to be engineered (which may be from a bacterium, a plant or a human) is put into a piece of viral DNA with the help of various technical tricks. The manipulated virus plus its fellow-traveller is then used to infect a new host. With luck, the recipient will treat the immigrant DNA as its own and make a copy every time its divides. It can be persuaded to generate vast numbers of duplicates of the engineered gene — and large amounts of whatever it manufactures; pure human proteins, drugs, or other materials.

To cross the sexual divide, deep as it is, between bacteria and the rest of life proved unexpectedly easy. Insulin was once extracted from the pancreas of pigs. The human gene was moved to bacteria and large quantities of the pure protein can now be made. Human growth hormone, too — once extracted with much controversy from the pituitary glands of the dead — is now made in the same way. This avoids a macabre and unexpected problem. A few patients caught a nervous degenerative disease from corpses that carried a virus. Now, the factor VIII gene, too, has been inserted into bacteria and patients are treated with its product.

Genetic engineering can also be used against infectious disease. Jenner could use the cowpox virus to vaccinate against smallpox (an experiment which would fall foul of the most lenient Ethics Committee today) because the viruses share antigens, cues of identity recognised by the immune system as the basis of its response. As a result, antibodies against cowpox protect against smallpox. Cow-pox itself can cause problems and even modern vaccines have a small risk of a reaction to the foreign protein. In any case, many diseases {such as leprosy) cannot be helped by vaccination because it is hard to grow their agents in the laboratory.

Some clever engineering gets round the problem. Antigen genes from an agent of disease are inserted into a harmless bacterium, avoiding the risk of infection as the genes for virulence have been left out. Antigens from several sources can be put into the same host to give a single vaccine against many infections. A modified strain of Salmonella {which in its native state can cause food poisoning) is used. The bacterium, with its added antigens, flourishes for a short time in the gut and, by persuading the recipient that he has been infected, ensures that antibodies are made.

Some of the tricks are simple. Plants can make copies of themselves from a few cells so that many can be produced from one without sex. It is hard to improve trees by breeding from the best, because it takes so long. Instead, a superior specimen has its tissues broken into single cells. Copies of that super-tree can then be grown to give, within a single generation, a super-forest. In the same way, natural vanilla, once extracted at great expense from a tropical orchid, has been replaced with the same chemical made by cultures of cells grown in the botanical equivalent of a factory farm.

The real promise for farming comes from inserting genes from one species into another. A certain virus causes what is almost a plant cancer: tissues lose their identity and the plant grows up distorted. This crown gall virus is good at picking up foreign genes and has been used to move them into new hosts. The first transformed plant, a strain of tobacco, was made in 1984, to great lack of public interest. A dozen years later, tomato puree made from engineered plants was on sale without much controversy. Then, though, public alarm began; and the 'Frankenstein Food' label was invented, gathering around itself a variety of cranks who claimed, with no evidence, that such foods were harmful to health.

Part of the problem is the word 'engineering', which sounds more of a threat than does the 'domestication' used of the first genetic manipulators. Part comes from the caution of biologists themselves. Thirty years ago they declared a moratorium (soon abandoned) on new experiments until safety rules were worked out. Most important, people are always suspicious of technical fixes; the idea that science can overcome all problems. From nuclear power to Concorde the optimism of engineers has often turned out to be short-lived. For the companies involved, public concern (helped by their own bland assurances about safety  and by simple arrogance in refusing to label engineered food) has proved a real problem. Monsanto makes many things (although it has now changed its name to disguise that faci); but became synonymous with a supposed attempt to poison the public. So alarmed is industry that it has set absurd standards of safety. One project used genes from Brazil nuts put into soybeans to provide a certain amino acid. As this is in short supply in the third world it might have saved thousands of children. Instead the project was abandoned as a very few people are allergic to the nut itself. The new plant might have killed one or two Americans a year. The end of the research was greeted as a triumph by the Greens. Other false accusations turn on the supposed dangers of resistance to an antibiotic, kannamycin, used to help pick out which engineered plants have incorporated foreign DNA. Kannamycin is not used in medicine, is widespread in nature, and its use in genetic manipulation is in any case becoming obsolete. Even so, kannamycin has been used as a stick with which to beat those keen to improve food production.