by Hamish G. Spencer
© Oxford University Press 2004 All rights reserved
Fisher, Sir Ronald Aylmer (1890-1962), statistician and geneticist, was born on 17 February 1890 at his parents' home in East Finchley, Middlesex, the last of seven surviving children born to George Fisher (1843-1920), a dealer in fine arts, and his wife, Katie (1855-1904), daughter of Samuel Heath, a London solicitor. He was the second of unexpected twins; the elder twin was stillborn.
Education and early influences
After a spell at Stanmore Park School Fisher attended Harrow School (1904-9) and won a scholarship to Gonville and Caius College, Cambridge, where he studied mathematics. It was here that an interest in eugenics was kindled, and he helped found the Cambridge University Eugenic Society in 1911. The following year Major Leonard Darwin, Charles Darwin's fourth son and president of the London-based Eugenics Education Society, spoke to the Cambridge group. Fisher was to form a strong friendship with Darwin, who encouraged Fisher's early scientific career both financially and as a mentor. Fisher, together with other members of the Cambridge society, acted as stewards at the First International Eugenics Congress held in London in July 1912 with Darwin as president. In the previous month Fisher had completed his degree as a wrangler with distinction, and won a postgraduate scholarship in physics for a further year at Cambridge.
After leaving Cambridge, Fisher worked as a statistician for the London-based Mercantile and General Investment Company before volunteering for army duty when the First World War broke out. His extreme short-sightedness, however, prevented any active service; he was almost blind without his thick glasses, especially later in life. Instead he taught at a number of public schools, but neither he nor his pupils enjoyed the experience.
On 26 April 1917 Fisher married (Ruth) Eileen Guinness (b. 1900), daughter of the preacher Dr Henry Grattan Guinness. The marriage produced two boys and seven girls, their number and upbringing a manifestation of Fisher's Church of England Christianity and his eugenic beliefs. The couple separated in 1943.
Fisher's first major contribution to genetics appeared in 1918. English geneticists at the time were embroiled in a debate between the biometricians (headed by Karl Pearson) and the Mendelians (led by William Bateson and Reginald Punnett) on the nature of inheritance. Fisher's paper effectively ended this debate by showing that the inheritance of continuous traits (such as height) studied by the biometricians could be fully explained by a Mendelian model of several genes acting simultaneously. The paper was also important for statistics: it brought the word 'variance' into statistics and laid the foundations of what later became possibly the most widely used statistical test, the analysis of variance. The work had been finished two years previously, but was rejected for publication by the Royal Society of London on the advice of the two referees, Pearson and Punnett. The paper was too genetical for the former, too mathematical for the latter, who even admitted that he did not understand much of it. Only the sponsorship of Darwin's Eugenics Education Society allowed publication in the Transactions of the Royal Society of Edinburgh. Fisher's paper came to be regarded as an essential part of the foundation of the science of population genetics, the study of the genetic constitution of populations of plants and animals, including humans. The synthesis of Darwinian evolution and Mendelian population genetic theory gave rise to neo-Darwinism, the prevailing view of evolution from the 1930s onwards. For the next twenty years Fisher, together with J. B. S. Haldane and the American Sewall Wright, established the mathematical basis of population genetics.
Rothamsted
In 1919 Fisher turned down a job under Pearson at University College, London, instead taking up what was to have been a temporary appointment as statistician at the Rothamsted Experimental Station in Harpenden, Hertfordshire. In the event, Fisher spent fourteen years at Rothamsted and his work there established him as arguably the most important statistician of the twentieth century. During his tenure he placed much of modern statistics on a firm theoretical footing and invented numerous now ubiquitous statistical tests: he ended the confusion about the number of degrees of freedom associated with Pearson's chi-square test, described the analysis of variance test and the concept of maximum likelihood estimation, and founded the principles of experimental design. Every modern statistician uses Fisher's work, most every day of their careers.
The argument about degrees of freedom was viewed by many at the time as a confusing and technical matter: which particular member of a family of distributions was the right one to use when assessing the statistical significance of Pearson's test? The debate soon became heated, in part because of Fisher's notoriously sketchy proofs, and also because he was often quick to take offence when he believed logic was being ignored. Fisher's solution revealed the fundamental importance of degrees of freedom, allowing the correct evaluation of many statistical tests in the future (in addition to the still widely used chi-square test). Fisher had already fallen out with Pearson because of the latter's cursory dismissal of Fisher's treatment of the distribution of Pearson's correlation coefficient, and the 1922 publication of his paper on degrees of freedom in the Journal of the Royal Statistical Society brought the conflict into the open. Fisher was claiming, after all, that Pearson did not understand his own test. Pearson responded in his own journal, Biometrika, but further exchange was limited when the Journal editors refused Fisher a rejoinder. He attempted unsuccessfully to obtain an explanation and in 1923, against Darwin's advice, resigned from the society in protest. Nevertheless, by 1926 it became clear even to Pearson that Fisher's geometrical insights had provided the correct solution.
The development of the analysis of variance is inextricably tied to issues of experimental design. The Rothamsted appointment was an attempt to wring more results from various completed agricultural experiments, but it soon became evident to Fisher that some ways of setting up experiments were intrinsically better at detecting possible consequences of the various experimental treatments. By 1925 Fisher had established the majority of the rules that all experimental scientists later came to use widely to ensure experimental efficiency. First, he emphasized the importance of replication in getting a reliable estimate of background variation ('error') so as to discount getting a certain result just by chance. The second of Fisher's design innovations was randomization: he argued that treatments should be allocated randomly to plots within a field, rather than in a systematic pattern, which was standard practice at the time. Randomization overcomes any unknown (and maybe unknowable) biases. Fisher's ideas, particularly randomization, were resisted at first, but the power of his methods soon became apparent to experimenters, who appreciated the importance of eliminating alternative explanations for their results. Fisher summarized his approach in Statistical Methods for Research Workers, which first appeared in 1925. It went through fourteen editions, the last from 1970, and was translated into six languages.
The statistical tests Fisher invented to accompany his designed experiments are collectively known as the analysis of variance, anova for short. In its simplest guise, an anova tests whether differences in a treatment (say types of fertilizer or strains of wheat) have different effects on the mean of some measure of interest (usually yield in Fisher's case). But Fisher extended it much further, showing how to extract the effects of two or more simultaneous treatments and, most importantly, their interaction. Such interactions may be very common and important, but they cannot be detected by the simple standard and apparently sensible approach of varying one factor at a time. In 1934 Fisher's chief publicist in America, G. W. Snedecor of the Iowa State College of Agriculture, gave the statistic now used in assessing the statistical significance of an anova the symbol F in honour of Fisher.
The other major statistical innovation from Fisher's Rothamsted tenure was the method of estimation known as maximum likelihood, which was propounded in 1922 in a sixty-page paper in the Philosophical Transactions of the Royal Society. The maximum likelihood estimate of a parameter, such as the true mean of a population, is that value most likely to give rise to the observed data. Fisher proved that maximum likelihood estimators have a number of novel but desirable statistical properties not necessarily possessed by other sorts of estimators. Maximum likelihood estimation became widely used in all areas of science, including, in the later twentieth century, the production of evolutionary trees showing the descent and relationships among various animal and plant groups. In recognition of his statistical advances Fisher was elected a fellow of the Royal Society in 1929.
Genetical work
Fisher's genetical work continued mostly at home during his period at Rothamsted and culminated in his 1930 book The Genetical Theory of Natural Selection, which he dedicated to Leonard Darwin. Much of the book was motivated by Fisher's eugenic concerns about the decline of civilization; the last five chapters were on human populations. Recent British census data showed that fertility decreased with increasing social status. Arguing that economic pressures favoured smaller families, Fisher was alarmed that the very men on whose skills civilization depended were siring fewer children. He proposed a system of family allowances under which fathers with higher incomes would receive higher benefits. Thus, regardless of social status, having more children would not be economically disadvantageous.
The book's lasting importance, however, comes from three of the earlier chapters, one about natural selection, a second about sexual selection, and a third about the way genetic systems evolve. In the first three decades of the twentieth century, most evolutionary biologists doubted the importance of natural selection as an agent of evolutionary change. Charles Darwin was credited with establishing much evidence for the occurrence of evolution but many biologists doubted his explanatory mechanism. Fisher's mathematical models, however, convinced him that selection was indeed a powerful force. In his book he showed that the rate of evolutionary improvement--or increase in mean fitness--in a population subject to directional natural selection was proportional to the level of genetic variation, a result he christened the fundamental theorem of natural selection. Subsequent work has shown the theorem not to be so fundamental, since it fails if many of its assumptions are relaxed, but it has been extremely influential in providing mathematical backing to Darwin's verbal description of how selection improves fitness.
Darwin's concept of sexual selection had also been ignored theoretically, and Fisher rehabilitated this idea as well. His description of how sexual selection could result in a runaway process, leading to exaggerated male secondary sexual characters, remains the most widely cited model of how sexual selection might act.
The ubiquity of the genetic phenomenon of dominance (in which one form of a gene, the dominant allele, expresses its phenotypic effect over that of a recessive allele when the two are together in a single organism) was accepted with little comment by most geneticists. Fisher, by contrast, wondered how dominance could have evolved. The third chapter of his book gives his view that modifiers of the level of dominance could have arisen through the action of natural selection. He had first proposed this idea in 1928 in a short paper in American Naturalist, which led to an increasingly acrimonious exchange with Sewall Wright, causing a lifelong rift between them. Wright argued that the selection pressure on such modifiers would be too weak. The debate reflected the two theorists' differing concepts of evolution. In Fisher's view by far the most important agent of evolutionary change was selection, whereas Wright had a more pluralistic approach, giving weight also to random genetic drift and the consequences of population structure. Fisher's long-term collaboration with Cambridge's E. B. (Henry) Ford was an attempt to document the power of natural selection in the wild. They met in 1923, while Ford was an undergraduate at Oxford, and Ford's magnum opus, Ecological Genetics, which appeared in 1964, is dedicated to Fisher's memory. Although Fisher's view about the evolution of dominance was not widely shared in later years, the idea that genetic systems could evolve remained.
University College and Cambridge
When Pearson retired from University College, London, in 1933, Fisher was appointed his successor as Galton professor of eugenics, while Pearson's son, Egon, was appointed to head a separate statistics department. Fisher made some attempts to heal the rift with the Pearsons, but the relationship remained strained, in part because of the overlap in subject and in part because of subsequent disagreements (involving also J. Neyman) about the validity of some of Fisher's experimental designs. Fisher continued his statistical and genetic work in this period, expanding his genetical researches to laboratory mice, purple loosestrife (Lythrum), wild primroses, and human blood groups, in particular the ABO and rhesus systems. Also at this time Fisher discovered that the results of Mendel's famous genetic experiments with peas were, from a statistical standpoint, too good to be true, igniting a controversy about how to account for the results that long continued. In 1936 he developed an important multivariate statistical tool, the discriminant function, which he illustrated using the now classic data on two species of flowering plants, Iris setosa and Iris versicolor, growing in a single colony.
In 1938 Punnett resigned as Cambridge Balfour professor, and, after a delay due to the Second World War, Fisher was appointed to the chair in 1943. He had now succeeded both referees who had rejected his 1918 genetics paper for the Royal Society. He continued his blood group, mouse, and Lythrum work, but the genetics department was not well supported. Resigning the Galton chair also meant losing editorship of the Annals of Eugenics and, together with the cytogeneticist C. D. Darlington, he founded Heredity, which they jointly edited. The first issue appeared in 1947 and it quickly became the leading British genetics journal.
Not all of Fisher's work was widely accepted. The outstanding example of an idea that seems to have failed is that of fiducial inference: Fisher's attempt over several years at avoiding the pitfalls he saw in Bayesian inference and the more traditional Neyman-Pearson sampling theory approach. Fisher matured his opinion in his 1956 book, Statistical Methods and Scientific Inference, but several philosophers and statisticians have since severely criticized it.
Final years
Fisher retired in 1957 and became a spokesman for the tobacco industry. A pipe smoker since his student days, he claimed that the link between smoking and lung cancer was not causal. This action was typical of Fisher, who enjoyed controversy. He travelled extensively in America before going to Adelaide, South Australia, which he enjoyed so much that he settled there in 1959. Fisher died unexpectedly on 29 July 1962 in Adelaide from an embolism after a successful operation for bowel cancer. None of his surviving children was in the country and his funeral took place at St Peter's Anglican Cathedral in Adelaide on 2 August without any family present. His ashes were buried in the cathedral next to a dedicated pew.
Fisher received numerous awards in his lifetime: election as president of Gonville and Caius College, honorary degrees from Glasgow, Calcutta, Chicago, and Adelaide, foreign memberships of the United States, Swedish, Danish, and Vatican scientific academies, honorary memberships of several international societies and various medals, including the Royal Society's highest award, the Copley medal, in 1955. In 1952 he was made knight bachelor by the new Queen Elizabeth.
Fisher was a prolific writer. Even in the year he died he published seven scientific papers. Over the course of his career Fisher wrote seven books, 294 scientific papers (reprinted in the five-volume Collected Papers of R. A. Fisher, edited by J. H. Bennett), more than fifty other smaller contributions (including letters to journals and obituaries), and hundreds of reviews of scientific books and papers, most of which appeared in Eugenics Review. Moreover, Fisher wrote on a wide range of topics: in addition to statistical, genetic, and eugenic works, his bibliography lists scientific contributions to, for instance, geometry and numerical methods, as well as statistical applications to meteorology, bacteriology, agriculture, soil science, and other fields. Selections from Fisher's numerous letters on evolutionary topics were published as Natural Selection, Heredity, and Eugenics (1983). One book-length biography exists, that by his daughter Joan Fisher Box. As one might expect, some of the debates in which Fisher was involved are given a more pro-Fisher cast (for example, the evolutionary role of genetic drift is deprecated) and there are several minor inaccuracies, but it covers his career and personal life exhaustively.
HAMISH G. SPENCER
Sources
J. F. Box, R. A. Fisher: the life of a scientist (1978)
F. Yates and K. Mather, Memoirs FRS, 9 (1963), 91-129
J. H. Bennett, ed., Collected papers of R. A. Fisher (1971)
R. A. Fisher, Natural selection, heredity, and eugenics, ed. J. H. Bennett (1983)
J. O. Irwin, G. A. Barnard, K. Mather, and F. Yates, 'Sir Ronald Aylmer Fisher, 1890-1962', Journal of the Royal Statistical Society: series A, 126 (1963), 159-78
Archives
Gon. & Caius Cam., MSS
Medical Research Council, London, corresp.
priv. coll., working papers and notebooks
U. Cam., department of genetics, working papers
University of Adelaide, corresp. and notebooks | Bodl. Oxf., corresp. with C. D. Darlington
Oxf. U. Mus. NH, letters to Sir E. B. Poulton
Rice University, Houston, Woodson Research Center, corresp. with Sir Julian Huxley
University of Sheffield, corresp. with Arthur Roy Clapham
Likenesses
W. Stoneman, four photographs, 1931-53, NPG
Elliott & Fry, photograph, NPG [see illus.]
photographs, repro. in Box, R. A. Fisher
photographs, repro. in Yates and Mather, Memoirs FRS
photographs, repro. in Irwin and others, 'Sir Ronald Aylmer Fisher'
photographs, repro. in Bennett, ed., Collected papers
[http://www.oxforddnb.com/view/article/33146]
GO TO THE OUP ARTICLE (Sign-in required)