You have four people over for dinner and serve a cream soup. Two are fine. One is bloated and heading for the bathroom half an hour later. The fourth waves it away: never touched milk, not since childhood.
Nobody here has a delicate stomach. Every mammal winds down the enzyme that breaks apart the sugar in milk once it is weaned. Winding it down is the factory setting. Drinking milk with no consequences is the exception.
So where does the exception come from? Not from the body on its own. That route opened the day somebody started keeping sheep and milking them.
The sugar in milk is called lactose, and an enzyme has to cut it apart before you can absorb it. In childhood that enzyme is switched on; around weaning it gets turned down. Turning it down makes sense: a weaned calf will never meet milk again, and keeping a machine running for nothing is waste. About two thirds of adults worldwide are still parked at that default setting, and too much milk leaves them bloated with a bad stomach. We call this "lactose intolerance" — but from the mammal's side, intolerance is the original edition.
In a minority of people the switch never closes. The cause is a tiny variant, and it does not sit in the lactase gene itself; it sits in a nearby stretch of sequence that works the switch. What changed is not the machine, it is the switch. In people carrying it, the enzyme stays on into adulthood.
What makes the case is that this happened more than once. Europeans mostly carry one variant; some herding populations in East Africa and the Arabian peninsula carry several others at different positions, each arisen independently, all doing the same job. The same pressure sifting out the same result over and over is the shape of the convergent evolution issue.
But do not reverse the order: the variant did not appear because milk needed drinking. When it appeared and where it landed had nothing to do with whether there was milk around. That part is chance. Where there was milk, the people who happened to carry it left more offspring on average; where there was none, it was a useless little change that came and went.
Then a genuinely counterintuitive layer. Milk fat residues in pottery show Europeans were milking and making dairy products for at least 9,000 years — more than 7,000 sherds from over 550 sites. Yet ancient DNA shows the variant that digests lactose became common thousands of years later than that. People drank milk for millennia before their bodies caught up even a little. The 2022 study that pulled this together adds an unexpected claim: the modest extra nutrition from milk cannot carry selection this strong, and the signal fits better once famine and pathogen pressure are in the model. On an ordinary day a bad reaction is just diarrhoea; in a famine year or a plague year, the dehydration can kill. That explanation is still being argued over.
Here is a heavier example. In parts of West Africa, the variant behind sickle-shaped red blood cells is unusually common. People carrying one copy face a lower risk of severe malaria and of dying from it; people carrying two copies have sickle cell disease, which is serious.
That distribution was not always there. In 1958 Livingstone tied it to farming: people cleared rainforest to plant yams, and once the canopy was open the ground held sunlit pools of standing water of every size. The mosquitoes that carry the malaria parasite happen to favour exactly that kind of pool. Their numbers rose, and malaria went from something you occasionally met to something permanently resident. Once it was resident, carriers of one copy left slightly more offspring on average, and the variant climbed through the population.
The cost has to be stated plainly. When two one-copy carriers have children, on average one in four gets two copies and a serious disease. So this is emphatically not "good for the population" — nothing is doing the accounting for a population. All that happened is that where malaria was resident, one-copy carriers left more offspring than non-carriers, so the variant became more common, and so did the suffering attached to it. "Evolution selected the sickle-cell gene to fight malaria" rolls off the tongue, but it turns a statistical outcome into an arrangement.
Details of the chain are still being filled in — which populations, at what dates, at what pace farming spread. But the format, practice changes the environment and the environment turns round and sifts people, holds up, and it has a name of its own: niche construction.
Bitter manioc will poison you if you eat it straight; it contains compounds that turn into cyanide inside the body. Some Amazonian peoples process it through an absurdly long sequence: scrape, grate, wash repeatedly, press, leave the pulp spread out for days, and only then cook it. Ask why those days of waiting are necessary and the answer is usually that this is how it has always been done. Yet when the cyanide is measured, every added step knocks the level down another notch, until what is left can be ignored. Nobody can state the chemistry, and the procedure is right anyway.
What really makes the point is what happened when it travelled. Manioc was carried to Africa, and the processing was shortened. In some places — especially in drought years and periods of upheaval, when the shortcuts get taken — an illness called konzo appears: the legs give out suddenly, and it does not reverse. In manioc's homeland, populations who have eaten bitter varieties for thousands of years show no record of that chronic cyanide poisoning. What went missing was not knowledge; nobody could explain the sequence anyway. What went missing was the sequence itself.
How did the sequence survive? Take it apart and the skeleton is the same as the genetic one: households differ in how they do it (variation), children copy the adults (transmission), and people who skip steps, along with their families, are likelier to get hurt, so others stop copying them (differential survival). What is being copied is not a stretch of sequence but a string of movements of the hands. Otherwise there is no difference.
Add a case somebody measured. In Fiji in 2010, researchers matched pregnancy and breastfeeding food taboos item by item against the risk that local fish carry ciguatera toxin: the forbidden species are exactly the large reef fish most likely to carry it, and keeping the taboos cut the chance of poisoning by roughly 30% in pregnancy and 60% while breastfeeding. Those taboos are learned from mothers, mothers-in-law and the older women a village recognizes as knowing — not worked out first-hand. One person testing alone could not accumulate that sample size in a lifetime.
Do not run this backwards, though. Surviving practices do hold knowledge nobody can articulate, but all we see are the ones that survived; useless and even harmful traditions travel just as long, as long as the cost of their failures does not land on the people passing them on. Age by itself is not evidence.
Genes have exactly one route: parent to child. Practices have many — parent to child, master to apprentice, colleague to colleague, one person to thousands who never met them. A new practice can spread without waiting for a generation to be born.
The nature of the variation differs too. Genetic variation is blind: where it lands has nothing to do with whether it helps. Cultural variation is often steered — you wanted the thing to work better, so you changed it and then passed it on. And transmission carries preferences: copy what most people do, copy whoever is recognized as good at it. The Fijian taboos travel on that second one.
Fidelity differs by orders of magnitude as well. Genetic copying errs very rarely; tell one sentence to ten people and you get ten slightly different versions, because the receiver rebuilds it in their own head rather than taking it down as written. That is why culture cannot be cut into clean units the way genes can — the sticking point for Dawkins' proposal that cultural units be treated as replicators. The working model now is dual inheritance theory: genes and culture as two parallel lines of inheritance, each with its own transmission rules, each changing the selection pressures on the other.
Lay the speed difference out and the conclusion arrives by itself: in coevolution it is almost always culture that moves first and genes that chase, and the chase does not close. On milk the genes caught up partly, over thousands of years, and only in some populations. Today's environment — sitting all day, screens lit until midnight, sugar within arm's reach — turns over far faster than that. The mismatch described in the issue on the evolution of emotion is therefore not a temporary glitch but the normal product of that gap, and as long as culture keeps accelerating the gap only widens.