A new story - the discovery that genetically engineered food damages the immune system - seems set to join the misleading half-truths so often paraded by the media in the campaign to convince us of the dangers of genetic manipulation.

The concept originated with a BBC Television World in Action program on August 10, which described experiments in which Arpad Puztai of the Rowett Research Institute in Aberdeen, Scotland, had apparently demonstrated that mice developed stunted growth and an impaired immune response as a result of eating genetically altered potatoes. Puztai said that he would not eat transgenic crops, at least until they had survived similarly exhaustive tests. Amplified by banner headlines in British newspapers and through television newscasts worldwide, World in Action's theme was the horror of "Frankenstein food."

There were, however, immediate grounds for caution over the practical significance of the Aberdeen work. The transgene concerned, taken from the jack bean, encoded concanavalin A, a lectin already known to be both immunosuppressive and toxic. And the results on potatoes were preliminary and unpublished.

The media claque set off by the program was not cautious. The Express ran a front-page splash, entitled "Genetic Crops Stunt Growth," accompanied by an editorial on Frankenstein food: "The latest revelations . . . raise the prospect that scientists might be creating something truly dreadful." The Daily Mail's front-page story announced that the discovery undermined repeated assurances from manufacturers and governments that such foods posed no risks.

Even the Daily Telegraph, which carried the news a day later, gave its page-one lead to the Rowett research and the government's rejection of demands for a ban on sales of foods made using gene transfers "despite the first example of one being found harmful to health." Meanwhile, other newspapers continued to campaign vigorously on the issue.

"There was a dread inevitability about the discovery that genetically modified food might pose a threat to our health," said the Daily Mail. "For the vast majority of us had already reached this conclusion, based on no greater expertise than the commonsense suspicion that scientists were busily tinkering in areas they did not fully understand."

But this was a furor that quickly ended in farce. Barely three days after World in Action dropped its bombshell, Rowett director Philip James revealed that the experiments had not been done using genetic methodology at all, but by spiking potatoes with concanavalin A. Arpad Puztai had been suspended. To its credit, the Daily Mail gave the same prominence to an announcement, "Food Scientist Got it Wrong," as to the original story. Other papers performed less creditably. The Express, for example, tucked the news into a single column with the ambiguous headline "Food Scare Expert Forced to Quit" - lending credence, perhaps, to the assertion of Sue Mayer from the pressure group Genewatch that this was "another step in a long line of stifling research workers from telling us what is going on."

So what are the lessons of this episode, with its forceful impacts on the feelings and fortunes of food producers, researchers, and consumers? First, just as scientists should not release data to the media until they have withstood appropriate critical scrutiny, so journalists in both electronic and print media need to pay far more attention to the question of whether and where research claims have been published.

This admonition may appear unnecessary and (in the era of the Internet) even unrealistic. Yet countless previous examples show that such prudence invariably serves the best long-term interests of scientists, the media, and the community at large. Although the refereeing process cannot guarantee absolute veracity, it undoubtedly helps to minimize error.

Second, journalists and editors should be careful not to expand from the particular to the general in a field of this sort. Many animal experiments over the past half century have brought to light toxic effects of specific antimicrobials. But such findings have, quite rightly, not been greeted with headlines asserting that antibiotics kill.

By the same token, it is daft to extrapolate even valid conclusions from one "genetically engineered food" to others, and to brandish this phrase as a generic shibboleth. There is little sense in using such a term to embrace even the handful of products already on the market.

Third, people commenting on genetic manipulation need to recognize that adverse effects coming to light during screening tests confirm, rather than repudiate, the effectiveness of those procedures. Even if the Rowett data were as they first appeared, there would have been no possibility of Arpad Puztai's hazardous tubers being approved for public sale. The Times was one of few papers to point this out in its short, calm, unprominent initial report.

With such rare exceptions, and despite the extraordinary disappearance of its raison d'être, this was not an impressive episode in the media coverage of science.