Why so much of science still judges results by a one-in-twenty chance
The famous p = 0.05 cutoff traces back to a 1925 handbook by Ronald Fisher. He called a deviation of about 1.96, or nearly 2, a convenient limit for deciding whether a result looked significant, since chance alone exceeds it only once in twenty tries. Researchers have leaned on that line ever since.
Fisher arrived at statistics through farming data. In 1919 he turned down a post under Karl Pearson in London and took a temporary job at the Rothamsted Experimental Station in Hertfordshire, where crop trials had piled up records since 1842. He stayed 14 years, and his 1921 study of crop variation was the first application of what became the analysis of variance. In the 1930s he framed experimental design around a lady who claimed she could taste whether milk had gone into the cup before the tea.
His reach extended well beyond significance tests. He created the modern method of maximum likelihood, built the z-distribution now used as the F-distribution, and in 1936 published the iris flower measurements that still serve as a teaching data set. With Frank Yates he described a pencil-and-paper shuffling procedure, driven by a table of random numbers, that programmers now call the Fisher–Yates shuffle.
In biology he did as much. A 1918 paper coined the term variance and argued that smooth variation in traits like height could arise from many separate Mendelian genes. His 1930 book The Genetical Theory of Natural Selection, dedicated to Leonard Darwin, helped found population genetics alongside the work of Sewall Wright and J. B. S. Haldane, and it revived interest in sexual selection. He liked to say natural selection generates an exceedingly high degree of improbability.
Born in East Finchley, London, on 17 February 1890, Fisher had such poor eyesight that he learned to solve problems geometrically in his head, and the army rejected him in the First World War. He won a mathematics scholarship to Cambridge in 1909. In 1933 he became Galton Professor of Eugenics at University College London, and because of that link several institutions, including his old Cambridge college, have since removed commemorations of him. He died on 29 July 1962.
Source: Ronald Fisher
