Page:Sm all cc.pdf/73

 The following two examples illustrate the challenge of identifying the causality that manifests as correlation: the investigators had to design experiments to tease out this causal pattern. In both examples, epidemiological studies of a large population were used to identify a statistical association between a pair of variables. What is the effect of electromagnetic radiation on health? In one study, pregnant women who used video terminals more than 20 hours per week had twice as many miscarriages as did other kinds of female office workers. The authors of the study cautioned, however, that radiation was not necessarily the cause of this difference. For example, the video-intensive jobs might be more stressful.

A statistical study of Denver children found that those who had lived near powerdistribution lines were twice as likely to get cancer than other children. This study was criticized for its uncontrolled variables, so other investigators conducted a follow-up study designed to be much better controlled and more diagnostic. Contrary to the researchers’ expectations, the new result was virtually the same as the original, so many scientists concluded that electromagnetic radiation really does seem to affect health. Note the origin of this change in opinions: the combination of a recognizably skeptical scientist and a tighter experiment [Stevens, 1992b].

Compelling scientific evidence is required, because of the potentially staggering human and economic impacts if a causal link between electromagnetic radiation and health were confirmed. A synthesis of more than one hundred studies demonstrates that health impacts are generally negligible [Derry, 1999], but scientific concerns persist, particularly regarding possible long-term effects of cell phones.

Is there a genetic predisposition to alcoholism? Research on this question exemplifies the problem of distinguishing between acquired and inherited characteristics. One of the most successful ways to attack such problems is by studying adopted children. For example, 30-40% of adopted children of alcoholics become alcoholics, compared to only 10% of the general population. This result constitutes good evidence for a genetic origin, but only because it was confined to children of alcoholic fathers; it is conceivable that an alcoholic mother could pass along an acquired dependence to her fetus, as occurs with heroin.

In a different type of experiment, H. Begleiter found a much higher incidence of certain deficiencies in thinking and remembering among alcoholics than among nonalcoholics. Some of these deficiencies disappeared after the subjects stopped drinking, but others persisted for years. Was this evidence of permanent damage caused by alcohol? The author considered a radical alternative hypothesis: instead of the brain deficiency being caused by drinking, it preceded the drinking and was a trait among those most likely to become alcoholics. In studies of children, he found that 30-35% of the sons of alcoholic fathers had the deficiency, although only 1% of a control group did [Kolata, 1992a]. Rare scientists (e.g., Bauer, 1994) claim that the continuing debates about acquired vs. inherited characteristics illustrate deficiencies of sociology. Many non-scientists interpret the debates as revealing the fallibility of scientists. Instead, this research exemplifies the inductive ingenuity of those scientists who can recognize the possibility of a pattern among incredible complexity, then design a test that successfully isolates the primary variables.