A handful of complementary procedures, however, stand out as different, and the most prominent of these is acupuncture. Where most alternative remedies have failed successive clinical trials, acupuncture has performed well, at least for pain relief.

Placebo-controlled trials have found positive effects against osteoarthritis, pelvic cramps, migraine and back pain. This week, a study from the University of Southampton compared acupuncture needles with fake ones that retract like stage daggers,

and found that the real thing produces different patterns of brain activity. This is the latest intriguing sign that it may be more than a placebo.

The picture is far from uniform. Even in pain relief, results have been mixed. A large German trial, published in the Journal of the American Medical Association on Wednesday, found that real acupuncture was no better than a sham version for migraine — though both were more effective than nothing. It is too much to claim, as the Independent on Sunday wrote of the Southampton study, that acupuncture is now “as good as drugs for treating pain”. But the results are worth following up, chiefly because there are credible scientific theories about how acupuncture might work. While homoeopathic remedies are without plausible medical function, it is reasonable to propose that inserting a needle into the body could provoke analgesic effects.

There are at least two good hypotheses that could explain this. The gate-control theory of pain, developed by Pat Wall and Ron Melzack in the 1960s, suggests that mild stimuli close a “pain gateway” in spinal nerves, preventing pain sensations from reaching the brain. Tens, a well regarded method of relieving pain with electrical stimulation of the skin, seems to

exploit this principle, and acupuncture may achieve the same thing. Another possibility is that small injuries such as needle punctures prompt the body to make endorphins, its natural painkilling drugs.

Either of these could illuminate apparently contradictory research, such as the two trials published this week. In the Southampton study, fake needles that did not pierce the skin were used as the placebo. The German one used real needles at non-acupuncture points.

The findings of both studies are consistent with the theory that sticking needles into the body has a painkilling effect and that placement according to ancient lore is not all-important.

This, though, is anathema to traditional acupuncturists. They insist that needles must be inserted precisely along “meridians”, bodily ley lines for the flow of a mysterious energy called “chi”.

The problem here, of course, is that neither meridians nor chi exists. As Robert Park, a particularly lucid debunker of pseudoscience, points out in his book Voodoo Science, this alternative anatomy was developed by sages who had never dissected so much as a frog. It takes no account of physiological understanding acquired by centuries of scientific investigation: the circulation of the blood, the germ theory of disease, the workings of the nervous system.

By holding to their theories, the more rigid advocates of acupuncture are doing their field a disservice. They invite rationalists to heap ridicule on a technique that may have much to offer, and hamper serious inquiry into its workings and benefits. If acupuncture can reduce pain, its biological basis needs to be understood to ensure that therapy is most effective. That is never going to be achieved without stripping out the mumbo jumbo.

Mark Henderson is the Times science correspondent