Scientists like to remind us not to confuse cause and effect. But they’re not immune from making that mistake themselves. Last week, for example, a flurry of sociological headlines emanating from a conference included the claim that elderly Taiwanese people who shop every day are 27% less likely to die over 10 years than those who shop once a week; and the claim that 16-year-olds who read books at least once a month are more likely to be in managerial jobs at 33 than those who read no books at 16.

It would be tempting but rash to conclude that shopping prevents death, rather than that ill health prevents shopping; or that reading causes career success rather than that a scholarly aptitude causes both reading and career success.

The nature-nurture debate has long been bedeviled by cause-effect confusion, as exemplified by the old joke: I’m not surprised that Johnny comes from a broken home; he would be enough to break any home.

Whole districts of Freudian theory are confused about cause and effect. For example, the incest taboo, forbidding people from mating with close relatives, turned out on closer investigation to be a codified expression of, rather than a cause of, incest avoidance. As Freud’s rival Edward Westermarck argued, there’s an innate tendency to develop revulsion at the idea of sex with close childhood contemporaries (who usually are siblings). A taboo turns this into a rule.

(For more on this see this book by Wolf and Durham.)

Nor is medicine immune. Some years ago epidemiologists found that women taking hormone replacement therapy had fewer heart attacks, but controlled trials found that HRT caused more heart attacks. It turned out that the women taking HRT in the epidemiological study were from higher socio-economic classes, so they ate and exercised better. Class caused both HRT and fewer heart attacks.

Even climate science has encountered cause-effect confusion. When in 1999 Antarctic ice cores revealed carbon-dioxide concentrations and temperature marching in lockstep over 400,000 years, many-including me- found this a convincing argument for attributing past climate change to carbon dioxide. (About 95% of carbon dioxide in the atmosphere is natural, coming from the exhalations of living things. In the past, carbon-dioxide levels rose as the earth warmed at the end of ice ages and fell as it cooled at the end of interglacial periods.)

Then four years later came clear evidence from finer-grained analysis of ice cores that temperature changes preceded carbon-dioxide changes by at least 800 years. Effects cannot precede their causes by eight centuries, so temperatures must drive carbon dioxide, chiefly by warming the sea and causing carbon dioxide dissolved in water to “out-gas” into the air.

(See this chart and the associated post by Frank Lansner at Jo Nova’s site:)

Climate scientists fell back on a “feedback” hypothesis, arguing that an initial change, probably caused by variations in the earth’s orbit that affect the warmth of the sun, was then amplified by changes in carbon-dioxide levels. But this made the attribution argument circular and left the reversal of the trend after a period of warming (when amplification should be at its strongest) still harder to explain. If carbon dioxide is still driving the temperature upward but it falls instead, then other factors must be stronger than expected.

Some climate scientists see cause-effect confusion at the heart of climate modeling. Roy Spencer of the National Aeronautics and Space Administration argues from satellite data that the conventional view has one thing backward. Changes in cloud cover are often seen as consequences of changes in temperature. But what if the amount of cloud cover changes spontaneously, for reasons still unclear, and then alters the temperature of the world by reflecting or absorbing sunlight? That is to say, the clouds would be more cause than consequence. Not many agree with Mr. Spencer, but it is an intriguing idea.