Muriel Bristol turned down a cup of tea over the order it was poured in. R.A. Fisher had drawn it from the urn and offered it to her. She told him she preferred the milk poured in first, and that she could tell the difference by taste. Not guess at it. Tell.

Bristol was a phycologist, a scientist who studies algae. The urn stood at Rothamsted, an agricultural research station in Hertfordshire, set out for the staff’s afternoon tea one day in the early 1920s. The station was built around a manor house, with a laboratory of bottled soil and crop samples (Figure 1).

A long brick manor house with curved Dutch gables, tall chimneys and a small bell turret, ivy over its middle, a low patterned brick wall and a lawn in front
An engraving of a long laboratory hall under a timber roof, its walls lined floor to ceiling with shelves of bottles and tables of figures, and five people sorting samples at a table in the middle

Figure 1. Rothamsted: the manor house in 1907 (left) and the laboratory in 1877 (right). Photo: Charles Latham, In English Homes (1907), public domain. Engraving: The Gardeners’ Chronicle (1877), Fig. 74, no known copyright restrictions. Both via Wikimedia Commons.

“Nonsense,” said Fisher, the station’s statistician (Figure 2). “Surely it makes no difference.” Bristol maintained that of course it did.

A young man in a dark suit and tie, small oval wire-rimmed glasses, his hair swept to one side, a medal pinned to his lapel
An older man with white hair and a full white beard, round dark-rimmed glasses, a dark suit and tie, looking off to the left in a sepia studio portrait

Figure 2. R.A. Fisher in 1913, at twenty-three, six years before he joined Rothamsted (left), and forty years later, in 1953 (right). Photographers unknown. Public domain, via Wikimedia Commons.

Then a voice from just behind them said, “Let’s test her.” It belonged to William Roach, a biochemist at Rothamsted who worked on insecticides.

Fisher and Roach set up the test at once, Roach helping with the cups. Bristol sorted them, and Roach declared her right.

Roach’s quick test could not rule out a lucky run of guesses. Fisher later worked out what it would take to rule that out: how many cups, and in what order. He published the answer in 1935, in The Design of Experiments. Its second chapter opens:

A lady declares that by tasting a cup of tea made with milk she can discriminate whether the milk or the tea infusion was first added to the cup.

Eight cups, four poured each way, presented in a random order. She is told there are four of each, and her task is to divide the eight into two sets of four.

The design is documented. The tasting is only remembered. The book never names the lady, never says where or when, and never reports how many she got right.

Bristol’s name, the place and the rough date come down to us through Joan Fisher Box, Fisher’s daughter and biographer. Box has Bristol getting more than enough of the cups right to prove her case. That is Roach’s verdict, from the man who called for the test and soon married Bristol. Her “personal triumph,” Box adds, “was never recorded.”

The familiar ending, in which the lady sorts all eight correctly, is in neither Box nor Fisher.

My husband claims he can taste aspartame. Given the diet version of anything, no label and no warning, he will name it before the second sip. He has had to say so more than once, and my answer was a softer version of Fisher’s. It is the same drink, I would tell him.

I was wrong in the way Fisher’s “Nonsense” was wrong. Aspartame is a different molecule and always was, the way milk poured first makes a different cup. Milk poured into near-boiling tea separates into drops, each meeting the heat long enough for the milk’s proteins to unfold, the way egg white turns white in a hot pan. That is the faint scalded note. When tea is poured over the milk, that is much less likely, and the cup comes out rounder. Eighty years after the tasting, the Royal Society of Chemistry gave that reason.

When I said it was the same drink, I was describing my tongue, not the drink. My tongue could not tell the diet drink from the original, and I was treating my tongue as the judge of his. The correction would have been easy: four tiny cups of the original and four of the diet, poured out of his sight. That would give his claim a way to come out right. I design experiments for a living, and I have never set out those eight cups for him.

Next Saturday morning, someone you live with takes one sip from a new bag of coffee, wrinkles their nose, and says it tastes burnt. You have drunk the same bag all week and noticed nothing, so you tell them it tastes the same. The answer takes one second. What would eight small cups prove that one sip cannot, before you both drink the rest of that bag? Neither of you pours a second cup. They stop mentioning the taste, and the next thing they notice, they keep to themselves.

What a Guesser Gets Right

Fisher’s design settles three questions before anyone pours the first of eight cups. A significance test, a check of whether a result could be luck, still answers the same three.

What gets believed for free. The starting assumption is that she is guessing, also called the null hypothesis. It is the one story that can be counted. No one at the urn could have said how many cups a real taster would get right: all eight, or six on an off day. But someone who cannot tell the difference still sorts the cups, and still gets a share of them right, with a probability that can be worked out in advance. That guesser is a mental model, a small-scale copy of how something works, and every count that follows is measured against it. Being able to taste the pouring order now means something specific: sorting more cups correctly than luck would.

Everything that could have happened. A guesser has a fixed list of ways the test could come out. She is told there are four of each, so every guess names four cups as tea first, seventy ways in all. The cups differ only in which was poured first, and the order is random, so a cup’s place in the row is independent of how it was poured. That makes the seventy equally likely. In part one, a hospital picked its nurse first and counted after. Fisher’s design shuffles the cups before anyone tastes.

Each guess fits a two-by-two table, cups poured against cups called. Every row and column adds to four, so one number fixes the table: how many cups she sorts right. Each tea-first cup she misses puts a milk-first cup in its place, so misses come in pairs: eight, six, four, two or none (Figure 3).

Eight cups in random order. The dark ones were poured tea first. Tap four cups to call them tea first; the rest are her milk-first calls.

Figure 3. One guess at eight cups, as a two-by-two table. Tap four cups to call them tea first. My own figure, built for this piece.

Four cups would not be enough. With two of each, a pure guesser sorts them all about 17 percent of the time. That is how often a die lands on six. A real taster could sort every cup correctly and still be told that luck manages that much. A four-cup test is a dismissal with a procedure attached. Six cups, three of each, bring a perfect sort by luck down to 1 in 20. Eight bring it to 1 in 70.

How rare the count would be under the default. The count is the number of cups she sorts right, a single number standing in for the whole test. Eight out of eight. That puts a pure guesser’s chance at 1 in 70. The p-value is that probability, about 1.4 percent: the chance of a result at least this extreme when the guessing story is true. It is not the chance that the guessing story is true.

p-value, n.

“Informally, a p-value is the probability under a specified statistical model that a statistical summary of the data (e.g., the sample mean difference between two compared groups) would be equal to or more extreme than its observed value.

P-values do not measure the probability that the studied hypothesis is true, or the probability that the data were produced by random chance alone.”

American Statistical Association, The ASA Statement on p-Values (Wasserstein & Lazar, 2016), “What is a p-value?” and principle 2

In the case of Lucia de Berk, a Dutch nurse, 1 in 342 million was offered as a p-value, built by multiplying three wards’ answers. A number that small is easy to hear as the chance that she was innocent, which is exactly what a p-value is not.

Counting the p-value straight from the two-by-two table is called Fisher’s exact test. Guesses pile up in the middle: 36 of the seventy, about half, get four right (Figure 4). Sixteen get six right. Just one gets all eight.

That is what one error would mean. One slip is two wrong cups, six right, and a guesser gets six or more 17 times in 70, about 24 percent. Eight cups leave the taster no room to miss one.

“At least this extreme” points one way. All eight wrong is just as rare, and would show she could tell the cups apart with the labels backwards. But her claim was that she could name them, so only high scores count against guessing: a perfect sort has a p-value of 1 in 70, not 2 in 70.

Figure 4. The seventy guesses, counted. The histogram gives each score’s chance and adds up the p-value; the arithmetic shows where each bar comes from. Built for this piece.

How small does a p-value have to be before the verdict goes to Bristol?

What Eight Right Cups Cannot Say

Fisher left that line to the experimenter, who can be “more or less exacting,” and in the same chapter named the usual choice: “It is usual and convenient for experimenters to take 5 per cent. as a standard level of significance.”

That line is called a level of significance. For Fisher it was a convention, one the experimenter could make stricter, and a p-value had to clear it before the evidence against the null counted. Eight right, at 1 in 70, clears 5 percent. A perfect sort of six cups, at 1 in 20, only ties it, and an experimenter who ignores results as likely as 1 in 20 learns nothing from it, which is why Fisher called a six-cup test useless. Neyman and Pearson gave the same name to a different line, written alpha: the highest rate at which a test may reject a true null, fixed like a pass mark before an exam is graded, so nobody can move it after seeing the scores. The likeness is only on the surface, Jose Perezgonzalez writes in a 2015 tutorial on teaching the two camps. Most textbooks in his fields, he adds, still teach the two as one blend, which he calls “very controversial.”

There is a second thing the p-value does not do, and a perfect score is what hides it. Eight out of eight makes it hard to go on believing she was only guessing. Six out of eight would have left the test silent, and a silent test says nothing about whether Muriel Bristol could tell the difference. Fisher said so in the chapter that lays out the eight cups: the null hypothesis “is never proved or established, but is possibly disproved, in the course of experimentation.”

Reading a silent test as a verdict is the ordinary mistake. A car’s new knock stays quiet for the one test drive at the garage, and the car comes back marked fixed. One quiet drive is a silent test, not proof the knock is gone. The question is how many drives the car had. Absence of evidence, taken for evidence of absence. So why do most retellings end on eight out of eight?

Why the Cups Keep Getting Poured

The woman who declined the cup has been written up again and again: in Fisher’s 1935 textbook, Box’s biography of him, David Salsburg’s popular history, and papers by Stephen Senn and John Richardson.

A claim meets skepticism, gets tested, and comes out vindicated. That pattern is satisfying whatever it is about: a coin called right ten times running, a psychic naming the card, a suspect picked out of a lineup. Told with a perfect score, this story is that trope, with statistics standing in as the excuse.

The trope needs a perfect score. Salsburg heard of one in the late 1960s from Hugh Fairfield Smith, a statistician Salsburg says had been there, and put it in print in 2001. Forty-five years from the tasting to the telling, thirty-five more from the telling to the page. In that version it is a Cambridge tea party, and the woman has no name.

Box, as Richardson quotes her, gives no count of the cups. Richardson, setting the surviving accounts against the records in 2021, ran Fisher’s eight through a second report from Roach, which George Box published in 1976: Bristol “made nearly every choice correctly.” Nearly every is not all. On eight cups, if she missed any, she missed two at least. Six or more right is what guessing produces about a quarter of the time.

Fisher himself, Maurice Kendall reports, said he never performed the experiment. Box has Roach helping with the cups, which leaves open who poured them.

Neither Fisher’s book nor Salsburg’s says who she was. Muriel Bristol held a doctorate and worked at Rothamsted. She published on algae and drew her own plates (Figure 5). A Danish phycologist, Johannes Boye Petersen, would go on to name a genus, Muriella, after her. In Salsburg she is one of the women at a Cambridge tea party, the claim hers and the experiment Fisher’s.

An engraved plate of branching chains of round and oblong algal cells, some packed with small spores, labelled A to I, signed B.M.B. del. at the bottom left and titled Bristol, Gongrosira terricola, n. sp.
A woman in a dark work coat bent over a microscope at a long bench of drawers, beside a tall bright window, glassware and lamps around her

Figure 5. Bristol’s hand and a Rothamsted bench. Left: a new species of alga, Gongrosira terricola, named and drawn by Bristol; “B.M.B. del.” at the bottom left means B. Muriel Bristol drew it, Plate II of her 1920 paper in the Annals of Botany. Right: Winifred Brenchley, a botanist at Rothamsted from 1906 to 1948, at her microscope. No photograph of Bristol seems to survive. Plate: public domain, via the Biodiversity Heritage Library. Photo: © Rothamsted Research, CC BY-SA 3.0, via Wikimedia Commons.

And her claim is worth a closer look. A claim like hers is a specific kind of claim: knowledge that lives in performance rather than explanation. Nobody argues their way to knowing whether Bristol could taste the difference. The same goes for my husband’s aspartame: neither my husband nor I can explain how he knows. The next time a partner or a child says the new coffee tastes burnt, asking how they know settles nothing. A shrug leaves the claim untested, and it goes on deciding what gets bought. The claim can only show up in a row of eight small cups, poured where the taster cannot see.

A Closing Reflection. Fisher’s cup at the urn stood for every claim someone makes about what their own senses pick up, and for the one-second answer that usually meets it.

  1. Doubt your one listen. Look at the earbuds or speaker nearest you. Does someone you know swear their new ones sound better than yours, when you hear no difference? If they want to pay more for the next pair, would your one listen be enough to talk them out of it?
  2. Recall the one-second answer. Tonight, think of the last time someone at your table or in your car said they could hear or feel a difference you could not: a new hum from the fridge, a new knock from the engine. What did you answer in the first second, and did they ever bring it up again?
  3. Pour the eight cups. This Saturday, before the next bag is bought, set out eight small cups for the person who made the claim, two coffees or two colas, four of each, poured where they cannot see, still steaming or fizzing, shuffled into a row. When they sip and sort, which result are you hoping for?

Fisher answered in one second. Roach answered with cups. This week a claim that makes no sense will land in front of you. Eight small cups, poured out of sight, are what it would take to find out.

Worth remembering Save as image

Your sip is a sample of one. It shows what your tongue can tell, not what is in the cup. So when someone tastes, hears or feels a difference you miss, test them before you say there is none.

Where This Came From

Joan Fisher Box does not name her informants for the afternoon, but her preface thanks William Roach for his recollections, and Roach, soon to marry her, had a stake in how it came out. Bristol died in 1949 by Richardson’s account; Senn (2012) gives 1950.

In Richardson’s account, the perfect score comes from Smith, through Salsburg. He finds Fisher with no Cambridge post then, and Smith at Edinburgh and Cornell. For two misses, Richardson gives p = 0.243. He concludes that Fisher was “apparently correct after all,” since the record does not show that she could reliably tell the cups apart.

“Absence of evidence, taken for evidence of absence” adapts a common saying: absence of evidence is not evidence of absence.

Intellectual Honesty Note. Fisher’s reply, quoted by Richardson (2021) from Box, is reported speech, not a transcript. Box has Bristol say only that the order made a difference; that she could tell it by taste is what the test took her to claim. The stake drawn from the marriage is my inference. “Quick test” renders Box’s word extempore. The scalded note and the rounder cup are my rendering of the chemistry, not a source’s. No photograph of Bristol seems to survive: the one some retellings print as hers was cropped from a staff photo taken after she left Rothamsted (Richardson, 2021), so Figure 5 shows a colleague, Winifred Brenchley, instead. The cup arrangement in Figure 3 and the kitchen test are built for this piece. The personal material is real.

References

Box, G. E. P. (1976). Science and Statistics. Journal of the American Statistical Association, 71(356), 791-799.

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

Bristol, B. M. (1920). On the alga-flora of some desiccated English soils: An important factor in soil biology. Annals of Botany, 34(133), 35-80. Public domain scan via the Biodiversity Heritage Library. https://archive.org/details/annalsofbotany341920balf

Encyclopedia of Life. (n.d.). Muriella J. B. Petersen 1932. Retrieved October 31, 2026, from https://eol.org/pages/90042

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

Fisher, R. A. [Unknown photographer]. (1913). Ronald Fisher in 1913 [Photograph]. Public domain, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:Youngronaldfisher2.JPG

Fisher, R. A. [Unknown photographer]. (1953). Fisher 1953 [Photograph]. Public domain, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:Fisher_1953.jpg

Gardeners’ Chronicle. (1877). Interior of the laboratory at Rothamsted [Engraving, Fig. 74]. The Gardeners’ Chronicle. No known copyright restrictions, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:The_Gardeners%27_chronicle_-_a_weekly_illustrated_journal_of_horticulture_and_allied_subjects_(1877)_(14782061044).jpg

Gigerenzer, G., Swijtink, Z., Porter, T., Daston, L., Beatty, J., & Krüger, L. (1989). The Empire of Chance: How Probability Changed Science and Everyday Life. Cambridge University Press.

Gill, R. D., Groeneboom, P., & de Jong, P. (2018). Elementary statistics on trial: The case of Lucia de Berk. CHANCE. Preprint: https://arxiv.org/abs/1009.0802

Kendall, M. G. (1963). Ronald Aylmer Fisher, 1890-1962. Biometrika, 50(1-2), 1-15.

Latham, C. (1907). Rothamsted, Hertfordshire, from the south-west [Photograph]. In In English Homes, Vol. 2. Country Life. Public domain, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:In_English_Homes_Vol_2_Rothamsted_Hertfordshire_from_the_south-west_31295007279283_0336.jpg

Perezgonzalez, J. D. (2015). Fisher, Neyman-Pearson or NHST? A tutorial for teaching data testing. Frontiers in Psychology, 6, 223. https://doi.org/10.3389/fpsyg.2015.00223

Richardson, J. T. E. (2021). A closer look at the lady tasting tea. Significance, 18(5), 34-37.

Rothamsted Research. (n.d.). Winifred Brenchley (1883-1953) [Photograph]. CC BY-SA 3.0, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:WinifredBrenchley(1883-1953).jpg

Royal Society of Chemistry. (2003). How to Make a Perfect Cup of Tea [Press release].

Salsburg, D. (2001). The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century. W. H. Freeman.

Senn, S. (2012). Tea for three: Of infusions and inferences and milk in first. Significance, 9(6), 30-33.

Sturdivant, R. (n.d.). Lady Tasting Tea [Course handout, MA376, quoting Box 1978]. United States Military Academy. Archived July 10, 2004, https://web.archive.org/web/20040710084649/http://www.dean.usma.edu/math/people/Sturdivant/images/MA376/dater/ladytea.pdf

Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: Context, process, and purpose. The American Statistician, 70(2), 129-133. https://doi.org/10.1080/00031305.2016.1154108

Wikipedia contributors. (2026). Muriel Bristol. Wikipedia. Retrieved October 31, 2026, from https://en.wikipedia.org/wiki/Muriel_Bristol

Wikipedia contributors. (2026). Ronald Fisher. Wikipedia. Retrieved October 31, 2026, from https://en.wikipedia.org/wiki/Ronald_Fisher

Wikipedia contributors. (2026). Winifred Brenchley. Wikipedia. Retrieved October 31, 2026, from https://en.wikipedia.org/wiki/Winifred_Brenchley