Copyright © 2005 the Brewery History Society

Journal Home > Archive > Issue Contents > |

## 'Student': William Sealy Gosset (1876-1937) Brewery Manager, Chemist and Statistician |

## by Philip Talbot |

William Sealy Gosset was one of the leading early statisticians who never intended to
become a brewery manager with Guinness. He was born in Canterbury, Kent in 1876 the
eldest son of the five children of Colonel Frederic Gosset of the Royal Engineers and
Agnes Sealy Vidal. The Gosset's were originally a French Huguenot (Protestant) family
who had fled to England in 1685 with the revocation of the Edict of Nantes, which
suspended religious toleration. The young William was sufficiently intelligent to gain a
scholarship between 1889-1895 at Winchester one of England's leading public schools,
which helped to alleviate the frugal existence of his family.^{1} Initially he attempted to emulate the career path of his father by entering the
Royal Military Academy at Woolwich to train as an officer in the Royal Engineers. Those
entering Wool-wich in this period were intellectually a cut above those entering the
Royal Military Academy at Sandhurst, the former requiring a sound scientific and
mathematical power of reasoning. Unfortunately for Gosset he suffered from severely
impaired eyesight and Woolwich ultimately rejected him.

Gosset then gained a scholarship to New College, Oxford where he achieved a first class honours degree both in Mathematical Moderations in 1897 and later in1899 in Chemistry. A career in academia appeared to beckon but he chose instead to join Arthur Guinness, Son and Co. Ltd at their St James Gate brewery in Dublin post graduation. It should be emphasised that Gosset was employed as a chemist, never as a mathematician or statistician, and he remained a Guinness employee for his entire working life. Horace Brown, who had previously worked at Worthington and was one of the most distinguished practical chemists of the era, headed the Guinness Research Laboratory that was opened in 1900 where Gosset was to work. The earlier work of the chemists Pasteur on the infection of beers and Hansen on pure yeasts had indicated a move towards a more scientific beer production process. Guinness under their inspirational Managing Director La Touche and in parallel with other leading brewers of the era such as Bass, Allsopp and Worthington had the financial resources to establish their own in-house chemistry departments. The scientific research undertaken by all these scientific parties was widely disseminated within the Institute of Brewing via its meetings and journal publications from its inception in 1886. The implication was that the brewing industry was inexorably advancing to a newer scientific basis, so much so that Charles Babington claimed in 1904, at the annual dinner of the Institute of Brewing, that,

To succeed now the brewer must turn to the laboratory. He cannot afford to neglect the teaching of science, as in the dark days of twenty-five years ago… Those days and methods are as dead and gone as King John and his sheet armour.

^{2}

However this enthusiasm belated reality as Sir Sidney Nevile observed in his
autobiography, *Seventy Rolling Years*; at this period scientists had not
made a widespread impact in the industry and traditional reliance on craftsmanship
prevailed, especially as it appeared to be delivering uncommonly good beer.^{3} Although the technical literature implied a modern scientific approach by the
brewery industry there remained a widespread indifference to science in a largely
conservative industry.^{4}

Guinness appears to be one of the notable exceptions to this conservatism. It was a
powerful example of an 'agribusiness monopoly' operating on a scale that came to almost
dominate the southern Irish economy so that by 1880 it was purchasing over half the
annual Irish barley crop.^{5} Guinness's progressive management had perceived the potential that science
offered to rationalise the traditional art of brewing and improve raw material
production. Thus the establishment of the company's Research Laboratory was driven by
commercial imperatives for obtaining the best and cheapest raw materials. The attainment
of these objectives included many variables such as the varieties of barley and hops,
conditions of dying, cultivation and maturing factors. Ultimately the company permitted
Brown and Gosset to research further the conditions of brewing, which allowed Gossett to
apply his mathematical skills. Gosset used the company's accumulated rich data on
materials, susceptibility to temperature change accompanied in conducting a series of
experimentation towards these ends.

In 1904 Gosset submitted his first statistical report to the Board of Directors on 'The Application of the Law of Error to the Work of the Brewery'. Gosset advanced the argument for further applying statistical methods, such as the error curve, random variables and hinted at the effects of correlation. However, he concluded that the existing standard methods of combining standard independent errors (later known as the law of distribution) were inapplicable in a brewery environment and that more advanced statistical methods were needed that were then beyond him. He had at the same time identified a further problem arising from large-scale brewing processes that prevented total control being exercised over the production process, which made any accurate experimentation impossible. Thus any subsequent statistically derived conclusions were necessarily probabilistic rather than certain. He stated,

… in such work as ours the degree of uncertainty to be aimed at must depend on the pecuniary advantage to be gained by following the result of the experiment, compared with the increased cost of the new method, if any, and the cost of each experiment.

^{6}

Gosset's first report had concluded that it was advisable to consult a professional mathematician about the degree of probability to be accepted as proving various propositions,

We have met with difficulty that none of our (mathematical text) books mentions the odds, which are conveniently accepted as being sufficient to establish any conclusion, and it might be of assistance to us to consult some mathematical physicist on the matter.

^{7}

This suggestion was taken up by Guinness and it led to an introduction in 1905 to one of world's leading mathematicians, Karl Pearson (1857-1936), a co-founder of the unique Biometric School, later the Department of Applied Statistics at University College, London, which pioneered new statistical techniques. Gosset developed a lasting friendship with the truculent Pearson and engaged in lengthy correspondence with him for many years afterwards. In their first meeting lasting ninety minutes, Pearson successfully conveyed to Gosset the principles of standard error theory, and Gosset returned to Guinness to practice these methods during the next year.

Guinness employed an enlightened system of sending staff away on specialised study. Gosset had already attended a course on brown beers at Birmingham University, where he was taught to use the haemacytometer, and had begun to study how to accurately measure the concentration of yeast cells. He was later allowed to pursue his education further by attending Pearson's laboratory, the Galton Eugenics Laboratory, University College, between 1906-1907 before returning to Dublin. During this period of study he worked on the Poisson limit to the binomial, the sampling distribution to the mean, standard deviation and the correlation coefficient. As Gosset realised

If the Brewery is to get all the possible benefit from statistical processes techniques valid for small samples had to be devised.

^{8}

At work Gosset was put in charge of the 'Experimental Brewery', which involved some
statistical work and he began publishing his statistical research in the mathematical
journal *Biometrika*. Publication of Gosset's work had required prior
permission from Guinness who had insisted on preserving company confidentiality by
excluding any company data from being revealed within these academic papers and by the
imposition of an author's pseudonym. Gosset chose to be published under the name
'Student' for which he his most usually remembered. This imposition was not unique to
Guinness or the brewing industry and the ban on staff using their own names was not
revoked until just before the outbreak of World War II. During this period Gosset once
more declined overtures to take up an academic post preferring to remain with Guinness.
Part of the reason was financial, Gosset was earning the then not inconsiderable sum of
£800 per annum, which was as now far above the academic salaries on offer and he had a
growing family to support following his marriage in 1906. Another reason was that Gosset
found the brewery work intellectually challenging and stimulating.

Gosset authored two papers *The Probable Error of a Mean* and
*Probable Error of a Correlation Coefficient* both in 1908 that both
dealt with the problem of small samples. This resulted in the creation of Student's t-
test for quality control in brewing, which Gosset claimed could draw inferences from
sample sizes a low as four. These techniques permitted a judgement as whether a series
of experiments, however short provided a result that conformed to any standard of
accuracy or whether further investigations were necessary.^{9}

At sometime around 1910 Gosset with the encouragement of Pearson began writing a book on experimental sampling in conjunction with the malster Edwin Beaven, who was an agent of Guinness. Only initial draft chapters were ever completed and Gosett's other work diverted him from ever completing the text that has not survived, and which he acknowledged would find difficulty finding a publisher. The intervention of the Great War also interrupted his work and Gosset attempts at more direct war work were thwarted by his poor eyesight and resigned himself to his brewery management work,

My own war work is obviously to brew Guinness stout in such a way as to waste as little labour and material as possible, and I am hoping to do something fairly creditable in that way. All the same I wish government would double the tax again, it's such an obvious waste of pig food now!

^{10}Gosset continued post war to work in Dublin although he found life difficult following the Easter Rebellion in 1916 and the ensuing Irish Civil War. In 1922 he was allowed a statistical assistant, Edward M. Somerfield, and later another assistant, A.L. Murray and thus began a distinctive small statistical department that he managed until 1934. At the same time Gosset gave papers at the Society of Biometricians and Mathematical Statisticians founded by Pearson and he also became a member of the Royal Statistical Society's Industrial and Research Section in 1934, to which he also contributed work.

In 1934 Gosset was given responsibilities over the new Guinness Park Royal brewery in northwest London and in 1935 he moved to the capital to become Head Brewer mainly concerned with scientific production. His tenure however was short-lived and in October 1937 he died unexpectedly of cardiac arrest.

Gosset was undoubtedly a major figure in the evolution of modern statistics, though his brewery background always undermined his position since both the brewing industry and mathematical academia remained suspicious of his allegiances,

To many in the statistical world 'Student' was regarded a statistical advisor to Guinness's brewery, to others he appeared to be a brewer devoting his spare time to statistics … though there is some truth in both these ideas they miss the central point, which was the intimate connection between his statistical research and the practical problems on which he was engaged … 'Student' did a very large quantity of ordinary routine as well as his statistical work in the brewery, and all that in addition to consultative statistical work and to preparing his various published papers.

^{11}

1. |
Plackett, R.L. and Barnard, G.A. (1990) |

2. |
Gourvish, T.R. and Wilson, R.G. (1994) |

3. |
Nevile, S.O. (1959) |

4. |
Richmond, L. and Turton, A. (1989) |

5. |
Vaizey, J. (1960) |

6. |
Mackenzie, D.A. (1981) S |

7. |
Plackett, R.L. and Barnard, G.A. (1990) |

8. |
Mackenzie, D.A. (1981) |

9. |
Ibid. |

10. |
Plackett, R.L.. and Barnard, G.A. (1990) |

11. |
Pearson, E.S. and Kendall, M.G. (eds) (1970) ‘William Sealy Gosset
1876-1937’. In: |