Unnatural Selection

or, how to build a new kind of creature with no foresight, no blueprint, and nothing in your hand but a sieve

Open your refrigerator drawer and you may be looking at one of the strangest facts in biology without knowing it. The broccoli, the cabbage, the kale, the cauliflower, the knobby kohlrabi, the little tower of Brussels sprouts — if you bought all six, you bought six versions of one plant. Not one family. Not one genus. One species: Brassica oleracea, a scruffy, bitter, unpromising weed that still grows wild on the chalk cliffs of the Mediterranean, looking like something you would step over on a walk and never name.

Nobody engineered those vegetables. No one sat down with the genome of wild mustard and a list of edits. Generations of farmers simply kept doing one small, stupid, obvious thing, over and over, for a few thousand years: save seeds from the plants you like best, and plant those. Some of them liked fat flower buds and kept saving the bushiest heads — that lineage drifted, century by century, into broccoli and cauliflower. Some liked big crinkled leaves and saved those — kale, collards, cabbage. Some liked a swollen stem — kohlrabi. Some liked the little buds that sprout along the stalk — Brussels sprouts. Same plant. Same starting deck. Six directions, because six sets of farmers had six different ideas about supper.

I am beginning here on purpose, and I am stealing the move from a better writer. When Charles Darwin sat down to drop one of the most disruptive ideas of the nineteenth century on a public he knew would recoil, he did not open On the Origin of Species with apes or fossils or the descent of man. He opened it with pigeon breeders. Domestic animals. Cabbages, near enough. He spent his first chapter on something every farmer in England already knew in his hands — that you can reshape a living thing by choosing who its parents are — precisely because it is undeniable. Start where the reader already agrees. Then walk.

So let us agree on the easy thing first, because it really is easy, and then let me hand you the controls.

The thing every breeder knows, stated plainly, is this: a population is not a fixed object. It is a cloud of small differences, and if you keep skimming the cloud from one side — breeding only from the tallest, the sweetest, the calmest, the bushiest — the whole cloud drifts that way. You are not adding anything that was not there. You are not designing. You are filtering. The variety is already rattling around in the population; the breeder just decides, each generation, which slice of it gets to become the next generation. Do that a hundred times and you have a greyhound. Do it a thousand times and you have corn.

Here is the part worth slowing down for, because it is the seed of everything this site is about. The breeder needs no foresight. The farmer saving the bushiest plant has no picture of broccoli in his head — broccoli does not exist yet. He just likes this one a little better than that one, today. The shape that eventually emerges was never anybody's plan. It is the accumulated residue of a million tiny preferences, each one local, each one blind to where the whole thing is going. A sieve, shaken again and again. That is the entire trick, and you can run it yourself right now.

The Experiment

Below is a patch of twelve wild plants, all cut from the same bitter weed. Pick a crop you would like to breed toward — the goal sits in the corner so you can see what you are aiming at. Then click the plants in the patch that look closest to it, the ones you would save seed from, and press Sow next generation. The children of the plants you kept will fill the patch, each a little varied from its parents. Keep skimming the cloud toward your goal. In a dozen generations or so, a bitter weed becomes a vegetable — and you never drew a thing.

Things to try:

What you just did has a name breeders have used for centuries — artificial selection — and the word “artificial” is doing less work than it looks like. The only artificial part is you, holding the sieve. The drift, the way small inherited differences accumulate into a new kind of thing when something keeps filtering them in one direction — that part is not artificial at all. That part is just arithmetic, and it runs whether or not anyone is watching.

Step out of the garden for a moment, because the real-world cases are wilder than any simulation.

Start with the dog asleep on your floor. A chihuahua and a Great Dane will not meet in the ordinary course of things, but they are the same species, and both of them are wolf — every breed, from the teacup to the mastiff, skimmed out of grey wolves over a span that is an eyeblink in evolutionary time. We took one predator and, by choosing who bred with whom, spun off herders and retrievers and lapdogs and sled-pullers, animals so different in size and shape and temperament that if you found their bones in the ground you would file them as separate species. One sieve, many hands, a few thousand years.

Or look at what is on the cob at a summer barbecue. Corn descends from a Mexican grass called teosinte, and teosinte is almost a practical joke at evolution's expense: a scrawny thing whose “ear” is a sad little spike of maybe a dozen kernels, each one locked in a shell hard enough to crack a tooth. It looks nothing like corn. It is so unlike corn that botanists argued for decades about whether it could possibly be the ancestor. It is. Mesoamerican farmers, saving the best seed year after year with no notion of genetics, walked that grass all the way to the fat sugary ear we cannot now live without — a plant so domesticated it can no longer reproduce without us, its kernels trapped on a cob that cannot scatter them. We bred it into needing our hands.

But the experiment that should give you a chill is a Russian one. In 1959, a geneticist named Dmitri Belyaev started breeding silver foxes on a Siberian fur farm, and he selected for one thing only: tameness. Each generation, he kept the cubs least afraid of a human hand and bred them. That is the whole protocol. Nothing about ears, nothing about coats, nothing about anything but nerve. And within a handful of decades — well inside a single human lifetime — the foxes did not just get tamer. They started coming out with floppy ears, curled tails, piebald white patches, shortened snouts, and a willingness to whimper and wag and seek us out, like dogs. Belyaev had grabbed one thread — fear — and a whole bundle of other traits came up tangled with it, because the genes that turn down a wild animal's alarm turn out to be wired to a dozen things nobody was selecting for. He watched a fox become something dog-shaped, in fast-forward, by pulling a single lever. You can go and read the records. It is one of the cleanest demonstrations we have that this is not a story about deep time at all. It can happen while you watch.

(That tangle — pull one trait and others follow — is the same lesson the genetic algorithm teaches on silicon, where selecting for a fast gait drags a whole posture along with it. And the selector need not be a person at all: in sexual selection, it is the peahen doing the choosing, breeding the peacock's tail a little longer every generation with no farm and no foresight, exactly as you bred the broccoli.)

I want to stop short of cheerleading here, because there is a place this kind of page can go soft — it can start to read as a hymn to the cleverness of breeding, as if every turn of the sieve were a gift. It isn't. The same hand that shaped the border collie, as breathtaking a working animal as most of us will meet, also shaped the bulldog that cannot breathe through its own crushed face, cannot whelp its own pups, cannot cool itself on a warm day. We selected for a look and dragged a lifetime of suffering up tangled with it, the way Belyaev dragged floppy ears out of fear — only this time the passenger was misery, and we kept breeding anyway because we liked the face.

The power has no opinion about what it is used for. Skim a population toward fast growth and you can feed a continent; skim a banana toward seedless sweetness and you get the Cavendish, a fruit so genetically uniform — every one a clone of every other — that a single fungus could erase it from the planet, the way an earlier blight already erased the banana your great-grandparents ate. Selection is a lathe, not a blessing. It will turn a thing toward whatever you point it at, including a cliff. That the universe permits you to shape life this way is not a promise that what you shape will be good. It only means the sieve works. What you breed toward is on you.

And that, finally, is the door this page is standing in front of without quite opening. You have just felt, in your own clicking hand, how little it takes to remake a living thing: variation in a population, inheritance from parents to offspring, and something — anything — that keeps deciding which ones get to reproduce. Three ingredients. In the experiment, the third one was you. Your taste was the sieve.

So here is the question to carry out of the garden. Everything you just watched happened because a chooser stood over the patch saying this one, not that one. Take the chooser away. Let nobody stand there at all. Let the cold and the drought and the hungry fox and the simple matter of which plants happen to leave more seed do the skimming instead — no taste, no goal, no hand. Does the drift stop?

It does not even slow down. That is the whole of it, and it is the next thing worth seeing: the sieve never needed a hand to hold it. But you do not have to take my word for that yet. You have a patch of wild mustard up above and a fresh afternoon. Go make a vegetable that has never existed, one click at a time, and notice — the longer you play — how little of you the process actually required.