American Scientist

The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century. David Salsburg. xii + 340 pp. W. H. Freeman and Co., 2001. $23.95.

Since about 1890, the study of statistics has reshaped much of social and natural science, along with important areas of medicine, business, engineering, policy and the law, without drawing much attention to itself. The public knows very little of the role of statistics, and even its practitioners are not accustomed to thinking of it as a historical force or as a product of history. There is by now a worthy body of scholarship on the history of statistics, written mainly by statisticians and historians of science, largely for an academic audience. David Salsburg aspires in this book to present statistics to a larger public and has structured his account loosely as a history. More precisely, it is a compendium of tales, the oral traditions of statistics. It relies very little on scholarly accounts, proceeding instead on the basis of what the author has picked up informally.

What should scientists expect of history? Some readers may recall a recent wave of editorials decrying the sad state of the humanities, unleashed by physicist Alan Sokal. Sokal submitted an article full of portentous obscurity and nonsense science to a small journal of cultural studies, which took his bait and published the piece without seeking expert advice from a physicist. The Lady Tasting Tea contains, in its first few chapters, errors so egregious and so elementary that we might almost imagine its author planted them to test the reviewing of a respected scientific publisher. Alas, nobody felt the need to consult anyone competent in history. The consequences are unfortunate.

"Until [Ronald Aylmer] Fisher, experiments were idiosyncratic to each scientist," Salsburg writes, implying that without the statistics of the last 80 years, scientific communities had no basis for judging experimental work. This is no mere factual mistake, but a historical absurdity. To illustrate the claim, he tells us that there was no good estimate of the speed of light before Albert Michelson's measurements in the late 19th century—which is untrue as to dates, misrepresents the point of Michelson's experiments (he wanted to detect the motion of the earth through the ether) and fails to support the point in question. The German doctorate in political science that Salsburg confers on Karl Pearson, founder of modern statistics, is a blunder, badly inconsistent with Pearson's actual background and training. We can more easily forgive his remark that Pearson changed his name from Carl to Karl out of admiration for Karl Marx, since that purely speculative suggestion by J. B. S. Haldane, who had no way of knowing, has taken root in the history books as if true. But Salsburg radically misrepresents the early decades of mathematical statistics when he contrasts R. A. Fisher, a dedicated eugenicist, with the Marxist Karl Pearson, "whose sympathies lay with the downtrodden." Pearson was in fact one of the founders of eugenics and was particularly acerbic in his eugenic views. Even the short biographical paragraph on him in my family's World Book Encyclopedia remarks on his eugenic commitments. Instances like these, and there are more, are not mere errors of detail, but elements of a historical account that is fundamentally untrue.

On more contemporary topics, Salsburg is less out of his element. He is not a historian, of course, but a statistician and a collector of stories. Unfortunately, he appears to have accepted and passed on uncritically what his interviewees told him, making no allowance for the notorious unreliability of memory and the possible unreliability of his sources. For example, his suspicions seem not to have been raised by Chester Bliss's account of life as a statistician in the Soviet Union in the 1930s. Bliss watched his laboratory supervisors "disappear" during the Stalinist purges and joined the Communist labor union for laboratory workers, yet he told tales of stalwart resistance to party meddling in his science, even to the point of mocking his interrogators with snide remarks about the gospel of Marx and Lenin. Salsburg serves this up as the story of a "naive" and "apolitical" scientist.

On statistical matters, we may suppose that Salsburg's training and experience have qualified him to distinguish what is valid or at least plausible from what is clearly erroneous. For purposes of a work like this one, which undertakes primarily to introduce the lives and work of some relatively recent statisticians, that kind of knowledge is almost sufficient. If someone had taken the care to clean up the historical material, his statistical knowledge and his love of anecdote would have sufficed to justify the book as an engagingly accessible romp through the modern world of statistics.

What he offers is, in many ways, well suited for a popular work. He goes lightly on the mathematics and even on occasion deprecates the abstractions of pure mathematics. Yet he provides nice insights into what statisticians can do with their analytical tools. He spent his own career at Pfizer, and he includes excellent material on statistics in medicine. He is interested also in uses of statistics in agriculture, industrial quality control, gambling, military work and censuses, to give only some examples. His points are developed through lively stories of the careers and the investigations of men and women in statistics. That is, he portrays statistics as a field practiced by interesting people and as a profession that has real significance for modern life. He does so in a way that educated readers without specialist knowledge will be able to understand and even enjoy. The book, when properly checked, may well merit a second edition. It did not deserve its first.—Theodore Porter, History, University of California, Los Angeles