ABSTRACT
As we know, the notorious ban on significance tests was instituted in 1999 in response to a groundswell of criticism of null hypothesis tests. Passions ran high in 1997, and gave rise to protest demonstrations in 1998 by students and some faculty at many institutions. The rallying cries of the activists were, “Support the total test ban,” and “Nix the Null!” There were rumors that at one point, militants had taken over the editorial offices of several journals, but this could not be confirmed. Due to the popularity of the cellular phone and the laptop, most journals are edited in airplanes, and the major airlines disclaimed any knowledge of hijackings by statistics students. You know the more recent history. There has been a gradual increase in complaints about confidence limits, hippogriff charts, and neo-Bayesian alternatives to significance tests. For some purposes, these methods produce too much information. Although it is doubtful that the pendulum of 2006 will swing all the way back to where it was in the last pre-ban years, serious consideration is being given to repeal of the ban on significance tests. In reflecting on statistical practice
over the past seven years, it may be helpful to consider how matters looked before the ban, when the issue began to be seriously debated. Recently, I ran across an expanded version of a talk delivered in June of 1996 at the annual convention of the American Psychological Society. I do not remember delivering such a talk, but the written text expanding on the talk is mine. In fact an original, shorter version, which stayed extremely close to the talk, was published under my name (Abelson, 1997) in the journal Psychological Science, with the fanciful title, “On the surprising longevity of flogged horses: Why there is a case for the significance test.” Here is the expanded version I discovered in my file, with a different title.
