At the start of the War on Terror, footage of dogs being killed by what appears to be nerve gas sickened people all over the world. Al Qaeda is evil, and what they did to the dogs was evil as well. But for every animal that al Qaeda has killed in this way the US has killed about ten million.

One of the most sickening experiments carried out by the US military involved strapping monkeys to platforms, which were designed to move like an airplane. The monkeys were taught to use a lever to keep the platform horizontal, and were given electric shocks when the platform went too far one way or the other. They were then subjected to nerve gas and irradiation until they were sick to the point of collapse. Rhesus monkeys have about the same level of consciousness and certainly the same ability to suffer as a three or four year old child. This experiment was continued one hundred times a day.

Dr Donald Barnes, who was at that time the principal investigator of the US Air Force School of Aerospace Medicine, calculated that 1000 monkeys were tortured in this experiment alone. The results were supposed to show how well Air Force pilots could keep going after being attacked by chemical agents. He concluded that not only were the experiments inexcusably cruel, they were also worthless: there is no evidence that any commander in the field ever used the data or even knew that it existed.

Apparently respectable companies that are quoted on the New York Stock Exchange carry out the vast bulk of animal experiments. For example, drug companies test new drugs on animals. This often involves creating an 'animal model': an animal such as a mouse, rat, cat or dog is deliberately bred to develop the disease. In the case of cancer, this means that these animals die a slow, lingering death. About 50 million animals die in drug experiments each year. Most of these new drugs do nothing new: they are simply slight variations on existing drugs.

In fact, animals are killed or tortured whenever any new product is put on the market. By law, the product's toxicity and irritancy must be measured using two standard tests: LD-50 and the Draize test. In the LD-50 test, groups of animals are given different amounts of the substance: the object is to see how much of it is needed to kill half of the test animals in a group. In the Draize test, the substance is put into the eyes of a group of rabbits. The animals are immobilized, and after a couple of weeks their eyes are examined. Depending on the substance, they might just be swollen, or else may have completely disintegrated. During this time, the animals are in constant pain.

Not only are these tests extremely cruel, they are bad science. The assumption is that results obtained from experiments with one animal are useful in determining how toxic the substance is in another. However, figures can differ by a factor of ten or hundred even between very similar species. For example, the test substance thiourea gives a value of 4 mg/kg in Hopkin's rats and from 1,340 to 1,830 mg/kg in Norwegian rats. It is probably safe to assume that Hopkin's rats are more closely related to Norwegian rats than either type is to humans. Results also differ depending on the time of year, the food that the animals have been eating, and the number of animals in the cage.

Dr Christopher Smith, a board-certified emergency medicine physician in the US, says that he knows of no instance in which an emergency physician has used Draize test data to aid in the management of an eye injury. Nor has he ever used LD-50 data to manage accidental poisoning. As he and many other toxicologists point out, it is much more useful to know symptoms, treatment and the results of clinical trials in humans. While it is certainly useful to know which substances are toxic, LD-50 values give a spurious air of precision to what is inevitably a very crude judgment.

Nor is the Draize test any more valuable. The FDA lost a major court case following an incident in which a woman splashed shampoo in her eye because they were unable to show that results obtained in rabbits were relevant to humans. Yet this was the only purpose that the test ever had.

So why do the tests continue? A central argument made by those people who support them is that it is better for a rat to die than for a human. This is debatable. In any case, the two are not even connected. If the data are not actually used, the only link is a psychological one.