Statistical Significance: Balancing Evidence Against Doubt

The notion of “statistical significance” is often used to quantify “reasonable doubt.” This paper briefly reviews the history of statistical significance. Scientific and legal approaches to the analysis of doubt are compared. Traditional scientific methods for describing error are shown to be comparatively stringent and inflexible. It is suggested that these attributes lead scientists to make firm value-judgments about the quality of their own evidence. Those judgments may be inapplicable in other situations. This process of pre-judging and censoring is surely undesirable when scientific evidence is presented in a wider context.