Gosset, William Sealy (‘Student’) (1876–1937), scientist, Guinness brewer, and applied statistician, was born 13 June 1876 in Canterbury, Kent, eldest of five children of Agnes Sealy Gosset (née Vidal) and Frederic Gosset, colonel in the Royal Engineers, who were married in 1875. Gosset was a scholar of Winchester College 1889–95. Due to poor eyesight, he was unable to follow his father into the Royal Engineers, and instead took up a scholarship at New College, Oxford, where he obtained a first in mathematical moderations in 1897, and a first-class degree in chemistry in 1899.
When Arthur Guinness, Son & Co. Ltd, manufacturers of stout in Dublin, wanted to incorporate new scientific methods into their brewing process and decided to hire some bright young science graduates, Gosset was one of the first they hired, and in October 1899 he moved to Dublin to take up a job as a brewer at St James's Gate. In his early years there Gosset became familiar with various aspects of the brewing process, and in 1904 wrote a report for Guinness on ‘The application of the law of error to work of the brewery’. The report emphasised the importance of the use of probability theory in order to set exact values on the results of experiments in the brewery, and in particular recommended that a mathematician be consulted about analysing results of such experiments where only small samples were being observed. This led to Gosset spending a year's leave of absence (1906–7) in the biometric laboratory of Karl Pearson at University College, London, where he obtained the statistical foundations for much of his later work at Guinness. During this period he was introduced to correlation coefficients and the large-sample theory which was extensively used by biometers at the time. He was aware, however, that modifications of the large-sample methods of Pearson would be necessary in order to deal with the special small-sample problems arising in the brewery.
William Sealy Gosset was a practical scientist, and published twenty-two scientific papers while working for Guinness. The research and methods developed were motivated by problems in the brewery dealing with the production of stout, arising in response to variations in barley, hops, and malt, and other experimental conditions. Gosset wrote under the modest pseudonym of ‘Student’. Arthur Guinness, Son & Co. Ltd was then at the forefront in using applications of scientific methods to brewing, and emphasised secrecy to deny important information to its competitors. In particular, the company was not keen on its employees publishing scientific research, and insisted they use pseudonyms.
Gosset rediscovered the Poisson distribution in his first publication, entitled ‘On the error of counting with a haemacytometer’, which appeared in 1906 in Karl Pearson's journal Biometrika (vol. iv, 351–60). In this paper, which was motivated by the problem of the distribution of yeast cells in a liquid spread thinly over a grid, he gave new practical applications for this now classic probability distribution. However, it was Gosset's second publication, ‘On the probable error of a mean’ (Biometrika, vi (1908), 1–25), which has proved to be a landmark paper in the history of statistics. Gosset, like many other scientists at the time, was interested in how the mean of a sample might vary about the mean of a given population in an experiment – in particular when the sample size is small. In this paper, he derived the probability distribution of the standardised version of the sample mean given by z = (x-μ)/s, and gave tables for this z distribution. In about 1922, R. A. Fisher, motivated by the need to address a wider class of statistical problems, suggested using the statistic instead of z, and ever since it has been known as ‘Student's t distribution’. It has been for years one of the most frequently used tools of the statistician. Another very significant piece of work (‘The probable error of a correlation coefficient’, Biometrika, vii (1908), 302–10) was his investigation of the probability distribution of the small sample correlation coefficient for a bivariate normal distribution. Gosset was always very interested in agricultural experimentation, and in his later work made many valuable contributions to the design and analysis of experiments. He was a man of very wide interests in spite of his heavy workload and dedication to Guinness. However, among all his activities and contributions, it is the Student t- test that has won Gosset a unique place in the history of statistics and scientific method.
He married (16 January 1906) Marjory Surtees Phillpotts, sister of another Guinness brewer, in Tunbridge Wells, Kent. He remained with Guinness for all of his working life, residing in Blackrock, Co. Dublin, until 1935, when he moved to London to take up his appointment as head brewer of the new Guinness brewery at Park Royal in north-west London. He died of a heart attack on 16 October 1937, and was survived by his parents, wife, children (one son and two daughters), and one grandson. R. A. Fisher, the great statistician and mathematician, wrote in his 1939 tribute to Gosset that ‘the untimely death of W. S. Gosset, at the age of 61, in October 1937, has taken one of the most original minds in contemporary science’.