 | ||
|
| |    How Social Scientists Prove Causality by Robert Needlman, M.D., F.A.A.P. reviewed by Laura Jana, M.D., F.A.A.P. Does child care cause emotional problems? Does eating broccoli prevent colon cancer? Does early reading aloud cause academic success years later? It's actually harder than you might think to prove that one thing causes another. The most straightforward way is to do an experiment. In an experiment, you change one pertinent factor in the test subjects' living situation-the amount of broccoli in their diet, for example--and watch what happens. If you can be sure that you are only changing that one thing, and if you get the same result time after time, then it's a good bet that whatever you changed caused the result. The problem is, when it comes to studying people, experiments are often unethical or unfeasible. For example, if you want to know whether spending 40 hours a week in out-of-home child care causes misbehavior, you can't simply pick a group of children at random and pack them off to daycare--their parents probably would have something to say about the matter! Nor could you simply force an unsuspecting group of grown-ups to eat broccoli five times a week. They just simply wouldn't do it. That's not to say the situation is hopeless, just complicated. Medical and social scientists spend their professional lives trying to overcome these sorts of difficulties. And even if you're not a scientist, it's still important to understand something of these complexities so you can decide for yourself whether a study that claims to "prove" that one thing causes another really delivers the goods. Be a discerning reviewer In weighing the evidence, scientists ask several common-sense questions. (1) Is there an association between the proposed cause and the proposed effect? That is, do they occur together more frequently than you might expect by chance alone? In order to decide how often you might expect two things to occur together, solely on the basis of chance, researchers use statistics. When things occur together more often than expected by chance, the scientists judge that the association is "statistically significant." To illustrate this, try this imaginary experiment: You have a bowl with 100 jelly beans in it; 50 are red, and 50 are black. Your scientific hypothesis is that red ones are "easier to pull out" (because you like the red ones!). You put on your blindfold, and pull out 20 beans. Lo and behold, you get 12 reds and 8 blacks. Does that mean that, indeed, the reds are easier to pull out? Of course not. The statistician in your study group calculates that, during any given trial, it's quite likely to pull 12 red and 8 black by chance alone. Too bad: Your findings are not statistically significant. Now imagine that you repeat your experiment 10 times, and in 8 out of 10 trials you pull out more reds than blacks. Now your statistician is impressed. The likelihood of this happening by chance is small, less than 1 in 100. Congratulations! Your hypothesis was supported: There is a statistically significant association between redness and being pulled out of the bowl. Your next step, as a good scientist, is to try to explain why that happens. Now here's a real and very current example: There is a debate going on about whether vaccination causes autism. On certain rare occasions, a child is vaccinated and two days or so later begins to show signs of autism. This might seem like a case of cause and effect but you have to understand the whole picture: Vaccinations are given very frequently during the years when autism typically first appears, so there is a great likelihood that autism and vaccination would occur one right after the other at least some of the time. According to most studies (and in the judgment of most experts), the frequency with which autism follows a vaccination is just what one would expect solely on the basis of chance. The conclusion is that there is no association between vaccination and autism. Without an association, it's very unlikely that one causes the other. (2) Are the results repeatable, especially by different researchers? The more the findings are repeated, the more confident everyone can be that they are real and not due to some unrecognized flaw in the study. It's rare that any single study proves anything, but several studies, all showing the same thing, often do. Almost always, though, there are one or two studies that seem to disagree. Sometimes they're poor studies, but other times there is an important reason that explains the disagreement and casts more light on the issue. For example, it might turn out that the studies that don't go along with the majority actually were done in groups of children who were different in some important way from the subjects of other studies-they were older, perhaps, or from a particular region of the country. Figure out the difference--and why it's important--and you've deepened your understanding of the entire issue. (3) Is there a plausible mechanism linking the proposed cause, and the observed effect? If I argue that being read to during infancy causes better reading ability in third grade (and I believe it does!), I have to be prepared to show what the steps are that lead from proposed cause to proposed effect, and I have to show that these proposed linking events actually take place. (4) Finally, is there an alternative explanation for the observed effect that might explain it as well or better than the one being proposed? In my experience, the answers to these questions are not easy to find in news reports about exciting new research that "proves" that one thing causes (or prevents) another thing. Many times, the reporters don't seem to have even asked the questions! So it's up to the audience to be skeptical, particularly about first "breakthrough" reports or studies that have never been done before. Some of these will turn out to be true, but many will end up being wrong. Since only time--and more research--will tell, it's wise in most cases to wait a while before making your own decision on the matter.
| |  | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| About Us |
Contact Us |
Our Partners Privacy Policy | Ethics | Advertising Policy | Terms of Service © Copyright 2004 The Dr. Spock Company. All Rights Reserved. THIS SITE DOES NOT PROVIDE MEDICAL ADVICE. The information drSpock.com
provides is for educational purposes only and is not a substitute for
professional medical advice. Always seek the advice of your health care
professional if you have a specific health concern. Mention or advertisement
of any product, service, or brand does not constitute endorsement, guarantee,
or recommendation by The Dr. Spock Company. Please read our full
Terms of Service. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
