The scientists also collected seeds from other weeds in the oilseed rape field and grew them in the laboratory. They found that two - both wild turnips - were herbicide resistant.

The five scientists from the Centre for Ecology and Hydrology, the government research station at Winfrith in Dorset, placed their findings on the department's website last week.

A reviewer of the paper has appended to its front page: "The frequency of such an event [the cross-fertilisation of charlock] in the field is likely to be very low, as highlighted by the fact it has never been detected in numerous previous assessments."

However, he adds: "This unusual occurrence merits further study in order to adequately assess any potential risk of gene transfer."

Brian Johnson, an ecological geneticist and member of the government's specialist scientific group which assessed the farm trials, has no doubt of the significance. "You only need one event in several million. As soon as it has taken place the new plant has a huge selective advantage. That plant will multiply rapidly."

Dr Johnson, who is head of the biotechnology advisory unit and head of the land management technologies group at English Nature, the government nature advisers, said: "Unlike the researchers I am not surprised by this. If you apply herbicide to plants which is lethal, eventually a resistant survivor will turn up."

The glufosinate-ammonium herbicide used in this case put "huge selective pressure likely to cause rapid evolution of resistance".

To assess the potential of herbicide-resistant weeds as a danger to crops, a French researcher placed a single triazine-resistant weed, known as fat hen, in maize fields where atrazine was being used to control weeds. After four years the plants had multiplied to an average of 103,000 plants, Dr Johnson said.

What is not clear in the English case is whether the charlock was fertile. Scientists collected eight seeds from the plant but they failed to germinate them and concluded the plant was "not viable".

But Dr Johnson points out that the plant was very large and produced many flowers.

He said: "There is every reason to suppose that the GM trait could be in the plant's pollen and thus be carried to other charlock in the neighbourhood, spreading the GM genes in that way. This is after all how the cross-fertilisation between the rape and charlock must have occurred in the first place."

Since charlock seeds can remain in the soil for 20 to 30 years before they germinate, once GM plants have produced seeds it would be almost impossible to eliminate them.

Although the government has never conceded that gene transfer was a problem, it was fear of this that led the French and Greek governments to seek to ban GM rape.

Emily Diamond, a Friends of the Earth GM researcher, said: "I was shocked when I saw this paper. This is what we were reassured could not happen - and yet now it has happened the finding has been hidden away. This is exactly what the French and Greeks were afraid of when they opposed the introduction of GM rape."

The findings will now have to be assessed by the government's Advisory Committee on Releases to the Environment (Acre). The question is whether it is safe to release GM crops into the UK environment when there are wild relatives that might become superweeds and pose a serious threat to farm productivity. This has already occurred in Canada.

The discovery that herbicide-resistant genes have transferred to farm weeds from GM crops is the second blow to the hopes of bio-tech companies to introduce their crops into Britain. Following farm scale trials there was already scientific evidence that herbicide-tolerant oilseed rape and GM sugar beet were bad for biodiversity because the herbicide used to kill the weeds around the crops wiped out more wildlife than with conventionally grown crops. Now this new research, a follow-up on the original trials, shows that a second undesirable potential result is a race of superweeds.

The findings mirror the Canadian experience with GM crops, which has seen farmers and the environment plagued with severe problems.

Farmers the world over are always troubled by what they call "volunteers" - crop plants which grow from seeds spilled from the previous harvest, of which oilseed rape is probably the greatest offender, Anyone familiar with the British countryside, or even the verges of motorways, will recognise thousands of oilseed rape plants growing uninvited amid crops of wheat or barley, and in great swaths by the roadside where the "small greasy ballbearings" of seeds have spilled from lorries.

Farmers in Canada soon found that these volunteers were resistant to at least one herbicide, and became impossible to kill with two or three applications of different weedkillers after a succession of various GM crops were grown.

The new plants were dubbed superweeds because they proved resistant to three herbicides while the crops they were growing among had been genetically engineered to be resistant to only one.

To stop their farm crops being overwhelmed with superweeds, farmers had to resort to using older, much stronger varieties of "dirty" herbicide long since outlawed as seriously damaging to biodiversity.