The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century

SALSBURG, D. The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century. W. H. Freeman, New York, 2001. xi + 340 pp. $23.95/L16.99. ISBN 0-7167-4106-7.

A lady tasting tea appears in the second chapter of R. A. Fisher's text The Design of Experiments (1935). An experiment is designed to test her claim to be able to tell whether milk was poured into each cup before or after the tea infusion. David Salsburg reports that H. Fairfield Smith claimed in the late 1960s to have been present, with Fisher, the lady and others, at a university tea party in Cambridge in the late 1920s at which this took place. This story starts Salsburg's book, and so his main title may be explained. As it happens, there is at least one other version of the story, recounted in Joan Fisher Box's 1978 biography of Fisher, who was her father. In this the lady in question was an algologist, Muriel Bristol, and the venue was Rothamsted. As Smith worked with Fisher in 1935-1937 and Fisher did not return to Cambridge as Professor until 1943, this other version seems the more likely original, despite the possibility that similar exchanges became, as it were, a favourite party trick of Fisher's.

What Salsburg offers in this book is a series of essays on statistics in the twentieth century, written in an easy and entertaining style. Most are centred on individual people who contributed to statistical theory or practice. Colourful, larger-- than-life characters and those whose careers included struggles against adversity are especially well represented. Fisher, Karl Pearson and Jerzy Neyman predictably head the cast, but many other players appear at least briefly. The approach is heavy on anecdote and will be popular as a readable source of statistical stories, some from the author's own experience. Someone has chosen cute or offbeat titles for the chapters, a few of which are very obscure or rather silly (`Advice from the lady in black', `The march of the martingales'). More importantly, the subtitle is inapt: the author seeks primarily to entertain and to inform, and his occasional quasi-philosophical passages do not add up even to showing that statistics `revolutionized' science, let alone explaining it. It is difficult to discern much logical or historical sequence to the essays, and there are a few repetitions of detail, but the book suffers little on either score, being one that can be dipped into according to whim.

Aiming at readers with little or no mathematical training evidently meant that the ground rules for writing were that there should be no equations, and even more strikingly no graphs or diagrams. (There are a few pictures, mostly photos of people.) On the whole, Salsburg did a good job of explaining many ideas within these limitations, but each informed reader will squirm at some of the vaguer or unduly simplified passages. The account of the bootstrap (p. 289) does not include any real explanation of the main idea, while John Tukey's early work on spectral density estimation is confused hopelessly with that nearly 20 years later on the Fast Fourier Transform (pp. 232-233). A more complicated issue is raised by statements (pp. 98, 111) that the terminology of p-values was used by Fisher. Is this true? Whatever the terminology, there is more emphasis in his work on reference significance levels of 5% and 1%, if only as a matter of convenience in working with published tables. The increasingly common use of p-values as a measure of weight of evidence against the null hypothesis is in fact closer to what Karl Pearson was doing much earlier, say in 1900.

The lack of in-text references is a frustration for the serious reader, especially when names and dates are given, but nothing beyond. A list of end-text references includes many splendid sources, but is skewed towards biographical pieces. Few of the most substantial works by historians of statistics are cited. Some of the more striking omissions are Constance Reid's biography of Neyman (1982), the biography of Student by E. S. Pearson, R. L. Plackett and G. A. Barnard (1990) and the essays of Stephen Stigler (1999).

What is most disappointing, however, is the high density of errors in the book, including checkable facts. Stigler has proposed a tongue-in-cheek principle of eponymy, that nothing is named after the person who really discovered it: Salsburg repeatedly misquotes this as 'misonomy', which would properly mean `hatred of laws' (p. 16, etc.). 'Anthropomorphic' for 'anthropometric' is another repeated malapropism (pp. 20-22). Weldon did not die in a skiing accident in the Alps, but in a nursing home in England (p. 21). To become a Wrangler in Mathematics at Cambridge is a splendid achievement, but it is quite wrong to say that were years in which no one was successful (p. 34). James Ware, rather than Arthur Dempster and Donald Rubin, is cited as co-inventor of the EM algorithm with Nan Laird (pp. 71, 85, 132). The biologist Peter Blackett should be the physicist Patrick Blackett (p. 92). Keynes' Treatise on Probability was not his Ph.D. dissertation (pp. 112, 305), and indeed he never wrote one. In the photo on p. 155, 'F. N. David' is Evelyn Fix, and vice versa. Fisher did not die on a ship returning to Australia, but in hospital in Adelaide (p. 180). Yvonne Bishop's book on log-linear models is quoted without naming her co-authors Stephen Fienberg and Paul Holland, or their mentor Frederick Mosteller: if ever a book was a joint product, it was that `green giant' (p. 206). The Institute for Advanced Study at Princeton is misnamed the Institute of Advanced Studies (p. 209). It is a myth that Tukey invented stem-and-leaf plots or box plots (p. 235): versions of both had been in the literature for some decades. Rather, what Tukey did was introduce catchy names and make people take notice of simple and useful ideas. Joan Rosenblatt, rather than Murray Rosenblatt, is cited as an inventor of kernel density estimation (p. 290). Beyond the examples here, there are dozens of other small slips. It is to be hoped that any reprint will be cleaned up somewhat.

References

REFERENCES

References

Box, J. F. (1978) R. A. Fisher: The Life of a Scientist. New York: John Wiley.

Fisher, R. A. (1935) The Design of Experiments. Edinburgh: Oliver and Boyd.

Pearson, E. S. (1990) 'Student': A Statistical Biography of William Sealy Gosset. Edited and augmented by R. L. Plackett with the assistance of G. A. Barnard. Oxford: Oxford University Press.

Reid, C. (1982) Neyman-From Life. New York: Springer. Stigler, S. M. (1999) Statistics on the Table. Cambridge, Massachusetts: Harvard University Press.