If modern medicine is simply the application of universal scientific knowledge to human bodies, then treatments for the same physical problems should be roughly similar all over the world. Instead, when Lynn Payer was researching her insightful book Medicine and Culture, she found that mastectomies were three times more frequent in New England than in England.

Then, too German speaking countries had three times the rate of appendectomies as other countries, while American surgeons performed six times as many coronary bypasses per capita than did their counterparts in Britain. As one doctor told her: "Plenty of people are still dying of diseases which other people do not believe in.''

Culture matters in health care, because physiological facts aren't all that go into defining a disease. A microbe may be identical all over the world, but the name, and nature, of the illness it causes are created by people. So fever and chills can make you wonder what's wrong with you, but the answer--swine flu, spirit possession, unbalanced chakras--will depend on the culture that formed you. Acknowledging this is not the same as claiming microbes, fevers and chills aren't real, or that prayers are as effective as penicillin against bacteria. It's just stating a fact about how people understand the world.

It is a consequential fact, Payer showed. If you're formed to think the heart is a mechanical pump, as Americans are, you're more inclined to open 'er up and fix the broken part. Not so if you see the heart as "an organ that has a life of its own, one that pulsates in response to a number of different stimuli including the emotions.’’ That, Payer wrote, was the German tradition, and it led doctors in Europe "to realize, long before Americans, that angina pectoris, or lack of oxygen to the heart muscle, can be caused not just by a clogging, but also by a spasm, of the coronary artery, a concept hard to integrate into the mechanical model. Pumps, after all, don’t go into spasm.’’

Now, if diseases of the heart show cultural variation in their definition and treatment, you shouldn't expect diseases of the mind to show less. Which is part of the point of Ethan Watters' new book, excerpted in last Sunday's New York Times Magazine.

Watters has been smacked for supposedly denying the physical reality of patients' suffering, and it's not hard to understand the impulse to defend psychiatry. If you've ever wrestled with mental illness--your own or an intimate's--then you know for a fact that crazy people don't feel eccentric or unconventional or a little bit further out on the same spectrum as the rest of humanity. They feel utterly apart, exiled from the sea of understanding in which the rest of us swim. You just know their condition's not a matter of perspective. Anyone who has despaired about this also knows the comfort of a label--of being able to find, in the categories of the DSM, a sense of certainty and a map of what to do.

If Watters wanted to strip people of their faith in their treatments, he'd be doing more harm than good. But his article doesn't say American psychiatric categories ought to be replaced by homeopathy or yoga. He simply notes that these categories come from a particular place and time. Yes, they are influenced by hard data about human brains. But they're also influenced by the politics of the American Psychiatric Association and the culture of American health insurance, which wants a code for every condition and prefers drug treatments to less quantifiable, more expensive approaches.

There are other ways of understanding a troubled mind. Worldwide, they're being crowded out, Watters argues, by American psychiatry's definitions. We shouldn't be so condescending and naive as to assume that this is bound to reduce the sum total of humanity's anguish. His thesis is that it's actually doing the opposite, by crowding out strategies that could work better. American psychology's claims to describe a universal human nature have been dented and dinged lately, as well they deserve. Watters is another blow, and it's worth reading in full.