A healthy colon holds a few hundred bacterial species — 150 to 400, at ten to the eleventh or twelfth cells in every gram of its contents. Nothing in that crowd is hunting anything. There is no recognition in it — no antibody, no receptor, no memory of a visit. No. 46 described the opposite arrangement: a room your body builds in a lymph node expressly to learn the shape of one invader. This is the defence that learns nothing, and for most of the pathogens you swallow it is the one that matters.
Its name is colonization resistance, and the usual one-liner — good bacteria crowd out bad ones — is not wrong so much as unhelpful. Crowding is the wrong noun.
What a full room means
The useful formulation is fifty years old and belongs to Rolf Freter. A newcomer that cannot glue itself to the gut wall has to eat, and to establish itself it must use at least one limiting nutrient better than every organism already using it. Not general scarcity — one resource at a time, each with its incumbent.
In 2021 a Munich group measured exactly that: a commensal E. coli blocked Salmonella Typhimurium almost entirely by drawing down galactitol, one sugar alcohol. The instructive part is their caveat — it only worked while the rest of the community was present, stripping simple sugars out of the gut. Alone, the same E. coli did far less. No member owns the defence, and it is not a sum of anything resembling a fight.
Nothing has to be killed
Here is the fact that reorganises the rest. The colon is not a container. It is a flow reactor, and everything in it is leaving — which is why peristalsis, the contraction rather than the flow, is the only reason anything stays at all. A resident holds its place only while it divides at least as fast as it is being washed out. No. 4’s sourdough starter holds its shape the same way, but there the persistence is the whole story; here it is the defence.
Apply that to something arriving: it need not be attacked, poisoned or identified. It only has to be held below the washout rate, and then it leaves with the contents. Which is why dose matters so enormously here: a small inoculum kept under the flush rate is not suppressed, it goes locally extinct. The residents do not win the fight. They make sure the plumbing wins it.
Worth separating from No. 120, also a defence made of numbers: cicadas work by satiation, too many of them to eat, where this works by subtraction, nothing left to eat. One overwhelms the consumer; the other starves it. And unlike the marram grass of No. 97, whose occupied ground turns against its own occupant, this occupancy is a problem only for whatever arrives next.
The room is not emptied — it is set
Now the inversion, and it is the best thing in the subject.
Your gut mucus is a sugar store, and certain residents — Bacteroides thetaiotaomicron among them — make a living cleaving sialic acid and fucose off it. In a healthy gut those freed sugars never accumulate, because other residents eat them the moment they appear. Measured levels sit near zero.
Antibiotics kill the eaters. The cleaving carries on. Free sialic acid and fucose spike — and C. difficile can catabolise sialic acid, though not fucose, while Salmonella will take either. The pathogen does not expand into a vacancy. It expands into a meal the survivors are still preparing.
So the defence was never a wall. It was an appetite, and you can remove an appetite without touching the kitchen. It is the tragedy of the commons run backwards: every resident drawing the shared pool to nothing is the parable’s catastrophe, and here it is the protection. The drug that rests the pasture is what starts the infection.
The mechanism everyone quotes, and the experiment that loosened it
There is a more elegant story than any of the above, and it deserves its fame and a caveat.
Your liver secretes primary bile acids, and one of them, taurocholate, is the signal that makes a dormant C. difficile spore germinate. Certain gut clostridia — Clostridium scindens is the celebrated one — convert those into secondary bile acids, deoxycholic and lithocholic, which are toxic to the growing cell. Read as a loop it is almost too tidy: the residents turn the invader’s own alarm clock into its poison.
In May 2025 a group in Bern tested that about as cleanly as it can be tested. They built two strains of a different 7α-dehydroxylating species, Faecalicatena contorta, differing in a single gene, and put each into germ-free mice. The working strain made secondary bile acids at C. scindens levels, and did not reproduce C. scindens’s protection. Same chemistry; no protection.
That is a preprint, not yet peer-reviewed, and it does not demolish the bile story — it uncouples it, which is worse for a tidy account than a contradiction would be. Something else strain-specific is doing the work, and the leading candidate is competition for proline. Food again.
What comes back
In 2018 a Copenhagen group gave twelve healthy men four days of three last-resort antibiotics, then watched for six months. The community recovered to near its starting composition in about six weeks — the reassuring half. The other half: nine species every one of the twelve had carried beforehand were still undetectable at day 180.
The pattern came back. Not all of the residents did.
≈
One loop I’m watching
Next: your cornea is clear for exactly one reason — a single layer of cells on its back surface spends your whole life pumping water out of it, against a tissue built to swell. Transparency is not a property of the cornea. It is a pumping rate. You were born with about six thousand of those cells per square millimetre, they do not divide, and you will never have as many again.