The 10 lines of the Halbridge experiment on day 13,715, each descended from the same ancestral strain of bacteria first thawed in 1988. Engraving by Ottilie Marsh
Every morning at 10, someone on the fourth floor of the old zoology building at the University of Halbridge takes 10 small glass flasks out of a shaking incubator, holds each one up against a white card printed with black stripes, and checks how cloudy it is. On the Tuesday I visited, in November, that someone was Noor Hadid, a 26-year-old doctoral student with a pencil behind one ear and a habit of reading each flask’s label aloud before she touches it. The flasks were cloudy, as they always are. Each held 10 milliliters of thin broth turned milky by several hundred million bacteria, every one of them descended from a single tube of Escherichia coli that a young lecturer named Hilde Marrack thawed on a Monday morning in May 1988.
What Hadid does next takes about 25 minutes and has not changed in its essentials since that first Monday. She wipes the bench with alcohol, lights a gas burner to keep the air above her work rising and clean, and lines up 10 fresh flasks of broth, made and warmed the afternoon before. With a new sterile pipette for each, she draws a tenth of a milliliter from every old flask and lets it run into its new partner: one part in a hundred. The new flasks go back into the incubator at 37°C, to be shaken gently for 24 hours. By tomorrow they will have eaten through their sugar and stopped growing, a hundred times more numerous than they began. Then someone will do it again. Hadid writes the day number in a bound ledger, with the time and her initials. The day I watched, it was 13,715.
Ilona Varga is the magazine’s science editor and writes about oceans, seeds, and experiments that outlast the people who start them. She spent 12 years as a marine biologist at the Westmere Coastal Station, counting fish larvae in plankton nets, before deciding she preferred describing patience to practicing it. Her book The Patient Sciences profiles eight studies older than their lead investigators. She lives in Westmere, swims off the north slipway in every month of the year, and keeps a sealed jar of runner-bean seeds on her desk.
Its telescope is a century old and its funding runs out in March. Its astronomers are using the time they have left to finish a sky survey begun in 1931.
Snapping shrimp, grunting fish and the crackle of life. Scientists are learning to restore reefs by playing them back their own noise.
Ilona Varga·7 min read
This article is free to read. You have 3 of 3 free articles left this month.Subscribe
Once a week she also streaks a drop from each flask onto a dye-laced plate. Half the lines carry a harmless genetic marker that turns their colonies red; the other half grow white. If a red line ever shows white colonies, something has crossed from one flask into another, and the morning stops being routine.
Marrack, now 74 and emerita, set out to test a question evolutionary biologists like to argue about over drinks: if you could rewind the history of life and play it again, would anything come out the same? No one can rerun the history of life, but you can do something smaller. She founded 10 populations from the same ancestral strain, gave them identical flasks, identical broth, and identical days, and set them going side by side to see whether they would evolve in the same direction or drift apart. She thought it would take three years to find out. “I wrote a grant for three years,” she says. “Nobody ever told me how to stop.”
A fossil record that wakes up
The experiment rests on a trick no paleontologist has. Every 60th day, before the transfer, Hadid draws off a sample of each of the 10 cultures, mixes it with glycerol, and puts three small vials of each into a freezer at −80°C. Bacteria frozen this way are not dead, only stopped. Thawed and given broth, they wake and grow. The freezer, a squat white cabinet in the corridor that hums a half-tone below the incubator, holds a record of the experiment at intervals of roughly 400 generations all the way back to the ancestor, and every layer of it can be brought back to life.
That means the descendants can meet their ancestors. Thaw a vial from 1988 and a vial from this year, mix them half and half in one flask, and let them compete for the same sugar. By the next morning, nine cells in 10 belong to the moderns.* Much of that advantage was won early, about half of it in the first 5,000 generations, the first two years or so, as the populations got used to life in a flask. But the curve has never gone flat. Each time the lab measures, the latest descendants are a little better than the ones before them, and nobody has a convincing theory of when, or whether, that will stop.
The lines have not all gone the same way, which is the most interesting thing about them. Four of the 10 have become what biologists call mutators: somewhere along the way they broke one of the systems that proofread their DNA, and now pick up new mutations roughly a hundred times faster than the others. In line seven, something stranger happened in 2003, around generation 36,000. The population split into two kinds of cell that still live side by side, one that eats the sugar fast and leaks a waste product, and another that has become good at living on the waste. Nobody designed it. It happened in a flask on a shelf, a day at a time, while people came in each morning to move a drop. Hadid’s doctoral project asks whether it would happen again. She has been thawing line seven at different points in its past and replaying it; so far the split has reappeared in three of 12 replays, all of them started after generation 30,000.
Two ledgers
Everything that goes into the freezer goes into a book first. The freezer book, now in its ninth volume, has a page for each freezing day, listing every vial by line, day, generation, box, and slot, so that anyone can locate the 1994 sample of line three without opening the door, because every opening costs the freezer some of its cold. There is a spreadsheet too, but the book is the authority. A second freezer, in the basement of the university library across the quad, holds a duplicate of every vial and its own copy of the book. The lab calls it the lifeboat. Between the two there are about 15,000 vials, and the lab can account for every one.
The other ledger is the day book, in which every transfer since 2 May 1988 is recorded with a date, a time, and a set of initials. Reading it is like counting the rings of a tree. There are the tight, careful entries of the first year, nearly all HM. There are summers of unfamiliar initials, when the regulars were away. And there is a Saturday in December 1997 with a ruled line across the page, after the incubator’s thermostat failed overnight and cooked the flasks at 49°C. The next entry reads, in Marrack’s hand: Day 3,520. All lines restarted from day 3,480 stocks. 40 days lost. Forty days was about 265 generations. The freezer gave them back everything else.
Who comes in on Christmas
Marrack retired in 2014 and handed the experiment to Tomas Achterberg, who had been her student in the 1990s and whose initials first appear in the day book in October 1993. She still comes in on Christmas Day. She has done the transfer on 35 of the last 37 Christmases, missing one for flu and one for a daughter’s wedding on the other side of the world, and is mildly put out that the lab now keeps a backup rota in case she cannot. “The bacteria don’t know it’s Christmas,” she says. “That’s rather the point. Someone has to be the kind of person who comes in anyway.”
The experiment has outlasted 11 grant renewals, four heads of department, a building refit during which the flasks were carried across the courtyard in a picnic cooler, and, by Achterberg’s count, 214 different sets of initials in the day book.† Most belong to graduate students who did the work for four or five years and moved on. The bacteria have changed a great deal in that time; the people have changed more often. If the experiment has a second subject, Achterberg says, it is the one nobody wrote a grant for. “We set out to measure evolution. We’ve also been measuring how long a group of people can agree to keep doing one small, dull thing properly.”
Hadid feels that weight more than she expected to. She checks the incubator’s temperature on her phone before bed. She works through the flasks in order, first to last, every time, and reads each label aloud, because one flask in the wrong slot could tangle decades of lineage. She has had, she admits, the dream in which she oversleeps and arrives at noon. She has her own way of describing the job: “The bacteria don’t need me to be clever. They need me to be here at ten.”
The bacteria don’t need me to be clever. They need me to be here at ten.
— Noor Hadid
She expects to finish her doctorate in 2027. By then the day book will have passed 14,000 entries, and NH will be one more ring in the tree. On Christmas morning this year, a Thursday, Marrack will climb the four flights to the lab—the elevator, she says, is for the flasks—light the burner, and move a drop from 10 old flasks into 10 new ones. Then she will write the day number in the book, and her initials beneath it, and go home for lunch. The day number will be 13,752. The bacteria will not notice who wrote it. The book will.
Notes
*Competition assays run by the lab in 2024 and 2025. The 1988 ancestor and samples from generation 88,000 of each line were thawed, mixed in equal numbers, and grown together for one day, with cells counted by plating before and after; marked and unmarked strains were paired so that the two could be told apart. Across the 10 lines, descendants made up between 84 and 93 percent of the cells at the end of the day, a mean of 89 percent, which corresponds to a relative fitness of about 1.7. The figure for the first 5,000 generations comes from repeating the assay with frozen samples taken every 5,000 generations. ↩
†The count was first made for the experiment’s 35th anniversary in 2023 and updated in November 2025. It treats each distinct set of initials as one person; at least two pairs of people are known to have shared initials, so the true number of people who have kept the experiment going is probably 216. ↩
