The current practice of feeding animals massive doses of chemicals to determine whether they might cause cancer in people has been attacked as misleading by two reports in today's issue of the scientific journal Science.

The existing system, which has led to federal bans of chemicals such as the artificial sweetener cyclamate and the apple-growth regulator Alar, considerably overestimates the risk of cancer for many chemicals, said Bruce N. Ames, a leading, though often controversial, biochemist and geneticist at the University of California at Berkeley.

"I don't think animal cancer tests are very useful for saying anything about human cancer," said Ames, who wrote one of the reports.

The problem, Ames said, is the test itself. Federal regulatory agencies such as the Environmental Protection Agency and the Food and Drug Administration have assumed that if a compound can cause cancer in animals, it can cause cancer in humans and that it raises the risk of disease no matter how small the dose. But for practical reasons, testers usually give animals huge doses of the chemical so that an effect will show up even in a small group of animals.

Scientists have long assumed that the chemicals cause cancer by damaging the animal's genes, causing cells to grow out of control.

That assumption is probably true but only for those chemicals that damage genes, said Samuel M. Cohen, a researcher from the University of Nebraska Medical Center in Omaha, who wrote the other of today's reports. Cohen said, however, there is growing evidence that a large number of chemicals that cause cancer in animal tests do not damage genes.

Ames believes that the doses of the chemicals being tested are so high that they exert a toxic effect that simply kills cells in various organs of the animal's body, stimulating surviving cells to undergo repeated division to replace the dead cells. Whenever a cell divides, there is a chance that its genes will be damaged and that the damage will result in a cancer-causing mutation.

It is akin to increasing the likelihood of something going wrong with a car's engine by sitting at a stoplight and flooring the gas pedal.

"Cell division is the key to mutation," Ames said. "You get mutations when cells are dividing, and the risk of getting a cancer is 10 times more in a dividing cell. A lot of cancer is due to cell proliferation," not the direct action of the chemical itself.

The implication, which could be profound for the way chemicals are regulated, means a lower dose of the chemical -- one nearer the real-life dose -- would not cause cell proliferation and, if it also caused no direct genetic damage, could not cause cancer.

One example, Cohen said, is melamine, an industrial compound that when given to mice and rats in high doses, causes the formation of stones in the bladder. The stones, in turn, irritate the bladder cells, stimulating them to divide repeatedly and ultimately leading to cancer. "If you give a lower dose of melamine to the rat or mouse, {a dose} that doesn't cause stones to form, you don't get cell proliferation or cancer," Cohen said.

Although some scientists and federal officials privately disagree with Ames, saying his research is sometimes inadequate, others say he is onto something.

"He is a widely respected scientist and we respect his ability, but that doesn't mean we are prepared to agree with all of it," said EPA Deputy Administrator F. Henry Habicht II. "We are confident that we are identifying synthetic chemicals that present a significant risk." Still, Habicht acknowledged "these tests are surely imperfect, though they are viewed by scientists as reasonable predictors of risk."

The EPA has asked the National Academy of Sciences to hold hearings on the best way to test chemicals for risks to humans. The first meeting will be held next week.