Jonah Lehrer has an article in a recent New Yorker about what can be called “the decline effect.” (I would link to the article, but it’s available by subscription only, and the available abstract is worthless.) Anyway, here’s the idea. Scientists from across the spectrum are finding that many of the correlations they’ve discovered seem to be “wearing off” over time. But why should this be? Lehrer suggests that it’s a matter of making big news over careful but flukey studies. So imagine this: scientists are always running all kinds of experiments and studies in many directions. Just as a matter of chance, some of them once in a while will “luck out” and examine a population which shows an especially strong correlation. Big news! Publication in Nature! Goes into textbooks! New nugget of knowledge! Then, as scientists try to replicate the study, they find weaker and weaker correlations. Hmm. The “decline effect” is especially troublesome in pharmaceutical research, where we’re finding that some of the fancy new drugs, despite initial appearances, aren’t all that more effective than drugs being used in the 50s.
But it’s hard to get these facts known, since journals are rarely interested in articles which report that scientists were unable to find some strong correlation. Somebody needs to go back and read some Popper, I think.