Review: The Lady Tasting Tea

David Salsburg’s The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century is a book that doesn’t know its audience. Billed as a history of statistical theories, it is more like a biography of the various statisticians who proposed those theories. Salsburg describes his intentions fairly well in the second to last paragraph of the introduction:

“Because statistical models of reality are mathematical ones, they can be fully understood only in terms of mathematical formulas and symbols. This book is an attempt to do something a little less ambitious. I have tried to describe the statistical revolution in twentieth-century science in terms of some of the people (many of them still living) who were involved in that revolution. I have only touched on the work they created, trying to give the reader a taste of how their individual discoveries fit into the overall picture.” (p. XI)

Popular science is of course always a balancing act: how much should you challenge your reader to understand difficult material? How much can you educate while still being engaging? Salsburg chooses to err on the side of simplification, spending lavish pages on the childhoods of various statisticians while summarizing their work in mere sentences. This is a valid choice, I suppose, though it drastically decreased my enjoyment of the book. And I can’t help wonder - who, besides people like myself with an active interest/education in statistics, would want to know the life stories of these people in the first place?

Salsburg’s approach is most obvious - and most grating - in the asides to the reader he makes throughout. On the very first page, he writes, “I can just hear some of my readers dismissing this effort as a bit of summer afternoon fluff… ‘There is nothing important or of great scientific merit in the problem’, they will sneer.” (p. 1-2) Later on: “The reader may recall those terrible moments in high school algebra when the book shifted into word problems… Imagine a word problem where nobody knows how to turn it into a formula… This is what happens when one tries to apply statistical models to real life problems.” (p. 75-76) Mostly, though, these asides can be ignored. The more fundamental problem is the shallowness with which Salsburg treats the statistics themselves.

His description of Jerome Cornfield’s work on the Framingham Study is a particularly egregious example. “Cornfield ran into fundamental problems that had not been addressed in the theoretical literature. Working with faculty members at Princeton University, he solved those problems. … Cornfield was satisfied to have found a method.” (p. 177) What method? What problems? To be fair, Salsburg usually provides the name of a statistician’s work, even if he doesn’t explain it. This at least lets the reader put down the book and search for a better understanding online. In my notes, I wrote down 19 different tests and terms I felt should have been better explained, from basics like coefficients of correlation, ANOVAs, goodness of fit tests, and degrees of freedom, to new-to-me concepts like contagious distributions, Kolmogorov’s statistics for stochastic processes, and martingales.

In another frustrating section, Salsburg describes the pioneering work of prominent statistician R. A. Fisher, whose story of a lady tasting tea provides the book’s title. Salsburg devotes a whole chapter (in this book, about ten pages) to a series of articles Fisher published called ‘Studies in Crop Variation’. This is actually one of Salsburg’s better chapters, where he describes in detail how the historical setting - in this case, an agricultural station trying to produce more and better crops - gave rise to various experimental and statistical methods - here, randomized controlled experiments and the now very common analysis of variance or ANOVA. Salsburg gives a decent summary of the first two articles in the series, but when he gets to number three - “a masterful article”, in which “lie the foundations of modern statistical methods used in medicine, chemistry, computer science, sociology, astronomy, pharmacology… highly ingenious methods of calculation… and many wise suggestions” (p. 50) - he does not give any clue as to the actual content of the article. Instead, he describes a photograph of Fisher, quotes the man on how some peoples’ feelings are hurt when their theories are disproved, and then ends the chapter. He never even gets to the remaining three articles. It’s up to the reader to look the articles up for themselves and figure out what they’re about.

Which leads us to Salsburg’s final major failing: this is a poorly referenced book. There are no endnotes or (Salsburg, 2001)-style citations in the text itself; instead, Salsburg nestles a ten-page bibliography at the back of the book. But he doesn’t list all his sources there. It is therefore not so easy to look up ‘Studies In Crop Variation’. As far as I can tell from googling, Fisher didn’t write all the articles in the series - not even the much-praised third article. The author on that article is actually one Winifred A Mackenzie. A further google search turns up nothing on this woman. Here is one case where I’m eager to know someone’s story, but Salsburg doesn’t provide it - he doesn’t even mention her existence. I find this somewhat ironic, given the pains Salsburg takes to let us know that Women Can Be Statisticians, Too. He has a whole chapter (again, about ten pages) describing the lives (though not the statistical achievements) of several women, ending with the encouraging: “Since the days of Snecedor and Cox, the ‘best person’ has frequently been a woman.” (p. 206)

This review is pretty negative, and that’s because I think the book is pretty flawed. That said, it does have some bright points. It lays out a wealth of information - people, articles, concepts, tests - that the reader can follow up on. And the moments where the author pushes his explanations and strikes the right balance of story and statistics are wonderful. The third-to-last chapter, on how clinicians dealing with the ethical issues of study design developed special methods, made it clear that Salsburg is capable of writing a fascinating book. This one, though, just isn’t it.