But set the insult aside, because underneath it is a real question, and it turns out to be one of the deepest questions in biology. For those of us who are smarter than a chimpanzee — what, precisely, is the difference? Not “we have bigger brains,” which is true and useless. I mean mechanically. You and a chimpanzee both start life knowing almost nothing and end it knowing a great deal. You both got from the first state to the second by learning. So the difference between you and the chimp is not that you learn and it doesn’t. It’s how. And the how is stranger, and more specific, than you think.
It begins with a paradox you can verify at any hospital.
The paradox in the maternity ward
A human being, the smartest animal that has ever lived — as far as we can tell, which is a hedge I’ll come back to — does not know what to do with its own baby. It has to be told. There are classes. There are books. There is, in the more affluent precincts of the species, a woman you pay by the hour to come to your home and explain why the baby is crying, whether the crying means anything, and whether your soul will ever fully re-enter your body. First-time parents stand over the crib at 3 a.m. googling whether that breathing pattern is normal. On this one task — the single most important thing any organism ever does, the thing evolution cares about with the cold intensity of a tax auditor — we are spectacularly, visibly incompetent.
Now consider a solitary wasp. It will die before its offspring hatch. It has a nervous system you could sketch on an index card. It never met its mother and never will. And it digs the burrow, hunts the caterpillar, stings it into paralysis rather than death so the meat stays fresh, seals the egg in with its living larder, and flies off to die — the entire program of parenthood executed once, flawlessly, with nobody to copy and nothing to learn. Zero training. Zero doubt. No 3 a.m.
So the conclusion writes itself, and it is the kind of conclusion that feels wonderful because it flatters us: every other animal just knows what to do with its young, and only humans have to be taught, and that is the clean bright line between us and the rest of creation. Instinct traded away for intelligence. The price of the big brain.
I believed a version of this for years. It is a lovely theory. It is also wrong — and, as usual, the way it is wrong is more interesting than the theory was.
The animal that should have made the case, and didn’t
Start with the counterexample a biologist reaches for first, because it looks like it should settle the matter. The theory says instinct is the crude, simple option and learning is the sophisticated one — so the most instinct-bound animals should have the simplest parental behavior, and anything elaborate should require teaching.
That is just false, and one bug is enough to show it.
The white-margined burrower bug, Sehirus cinctus, guards her eggs and then forages to bring seeds back to her newly hatched young — real provisioning, adjusted to how hungry the brood is (documented maternal behavior). It is genuinely elaborate. It is also not taught. She is not handing down a recipe — not listen carefully, Brenda, the mint seeds are best in late May. The behavior is species-typical, built into the bug’s inherited developmental program, run more or less the same way by every female of her kind whether or not she ever watched another do it. Which makes the useful point, the one to carry forward: elaborate behavior is not the same as learned behavior. Complexity does not imply a teacher. A great deal of impressive parenting, across the whole animal kingdom, is simply run off the genome — no watching, no instruction, no tradition. So the question is not whether behavior is complex. It is where the information lives.
Now swing to the animal nearest us: the great ape. If the human helplessness at the crib were really the price of our particular intelligence, it should be ours alone — a gorilla ought to have kept the instinct we supposedly traded away, and mother competently the way the wasp does.
It doesn’t. A great ape is as helpless with its first infant as we are with ours, and for the same reason: it never learned. A captive gorilla raised without watching other gorilla mothers frequently cannot mother at all — it won’t hold the baby, won’t nurse, has to have the whole thing taught to it. This is common enough that zoos run programs for it. The Smithsonian’s National Zoo has coached an expectant mother through holding and nursing using photos of gorilla mothers and a plush infant that plays newborn calls (their account); the Cincinnati Zoo has gone further, putting human surrogates in faux-fur vests to carry rejected infants, vocalize like gorillas, and teach the baby to cling as a gorilla mother would (documented by the zoo). We are, and I want you to sit with this, running classes for gorillas on how to be a mother — the same remedial course we run for ourselves.
So the gorilla is not a case below the human; it is the same case as the human. Both are big-brained, elaborately capable animals whose mothering is not shipped with the body but has to be picked up from other members of the species — and both fail, identically and helplessly, when it isn’t. (A hand-reared ape is damaged in many ways at once, not just maternally, so no single case settles exactly how much rides on learning. But the pattern is not subtle, and it repeats.) The bright line we were promised — humans who learn on one side, animals who just know on the other — has quietly stopped falling between the species and started falling inside them: between the individuals who got to learn and the ones who didn’t.
And the gradient runs down through the mammals. First-time mothers across a huge range of species are measurably worse at the job — less milk, more stillbirths from botched deliveries, higher infant death — and they improve with each birth. Not because the mother got smarter between pregnancies. Because she got experience. The competence that looked like a birthright is, across much of the mammal line, at least partly a skill.
Which means the wasp and the gorilla are not two marks on one ruler. They are running two different machines. And that is the question worth an essay — the same question we started with, sharpened to a point:
When a species could transmit knowledge by learning — when it lives long enough, and is smart enough, that teaching its young is genuinely on the menu — does evolution take the option? Or does it keep using the genome anyway?
Where the information lives
To answer that, stop thinking about babies for a moment and think about information — the stuff a parent has to get into a child: what to eat, what to fear, how to handle prey that fights back, how to raise the next one in turn.
Behavioral information reaches an animal by three broad routes. Some is built in by natural selection — written into the developmental program, arriving with the body, no experience required. Some is acquired by individual learning — the animal works it out itself, by trial and error. And some arrives by social learning — from other individuals, by watching them or being shown.
Those routes differ in cost and in reach, and the differences decide everything.
The built-in route is a magnificent medium with one crippling limitation. It is reliable, it is instant, it works even if the parent is dead before the child hatches — the wasp, the burrower bug, the monarch butterfly that navigates from Canada to a grove in Mexico it has never seen, on sun and magnetic cues wired in before birth. But it cannot be updated within a lifetime. You get the program your ancestors got. If the world has changed since they wrote it, you lose.
So the moment a species lives in a world that changes faster than the genome can track — shifting prey, new predators, new terrain, other minds — it has to move some of the information off the genome and into the living animal, through learning. Individual learning is one way, and it is cheap for everyone except the learner: a New Caledonian crow works out in a water-displacement task that dropping stones raises the water and floats a treat within reach, with nobody to show it — but trial and error is slow, and lethal when the first mistake is the last one.
Social learning fixes that by letting the naive animal skip the dangerous trials and crib off someone who already knows. Most of it is cheap for the knowledgeable animal too, because it isn’t doing anything extra — it’s just living its life in view. A wolf cub learns the hunt by tagging along while the pack hunts; the pack isn’t teaching, nobody slows the elk down for pedagogy, the hunt simply happens in front of the cub and the cub picks it up.
And then there is the expensive corner of social learning — teaching. This is the one case where the knowledgeable animal pays: it changes its own behavior, at a cost or for no immediate benefit, specifically so the naive one learns faster or more safely. That single fact — that teaching costs the teacher — is the hinge the whole question turns on. Because a cost is exactly the kind of thing natural selection refuses to pay unless it has to.
Nature is a tight-fisted CFO
So: a species that could teach — long-lived, iteroparous, young underfoot for years, a brain plainly up to the task. Does evolution reach for teaching?
Overwhelmingly, no. It leaves the capacity on the shelf and pays out of a cheaper channel.
Here is how we know, and it’s a genuinely startling fact. Biologists have a strict operational definition of real teaching — the Caro-Hauser criteria: the knowledgeable individual modifies its behavior only in the presence of a naive one, at a cost or with no benefit to itself, and the pupil learns faster as a result. Apply that definition honestly across the animal kingdom and the confirmed teachers are not the roll of brainy, long-lived species you’d expect. The classic experimentally-supported cases are meerkats, pied babblers, and tandem-running ants, with a thin and contested tail of proposed others; reviews still treat unequivocal evidence for teaching as scarce next to the far wider spread of ordinary social learning, and researchers keep arguing over whether the definition is too strict or exactly strict enough. Meanwhile elephants, wolves, most primates, the corvids — long lives, big brains, offspring around for years, every prerequisite satisfied — mostly do not show rigorous teaching at all. The capacity is everywhere. The use is rare.
And now the detail that flips the table. Look at who is on the list. An ant. A meerkat. These are not the intelligentsia. Teaching, it turns out, does not require a big brain at all — Alex Thornton, who ran the experiments proving meerkats teach, likes to say the whole point is that meerkats teach despite not being clever. So intelligence is not the switch that turns teaching on. Cognitive capacity opens the door; evolutionary economics decides how often anything walks through it.
Teaching evolves only when the young animal cannot get the skill any cheaper way — when the genome can’t carry it, watching can’t convey it, and trial-and-error would kill you before the lesson landed. Only when all the cheap channels are exhausted does evolution, grudgingly, pay for the expensive one.
Watch it happen in the one case that clears the bar. A meerkat pup has to eat scorpions. Scorpions have a stinger that can kill the pup. You can’t hardwire scorpion-handling — live scorpions vary too much to reduce to a fixed program. You can’t leave it to trial and error — the first error is the last. And watching an adult do it at full speed is no use to a pup that isn’t ready. So the adult does the one thing left: it bites the sting off, hands the disabled scorpion to the pup to practice on, and makes it progressively harder as the pup improves. That is teaching in the full technical sense, performed by an animal nobody would call bright, purely because for that one lethal skill there was no cheaper channel. The instant a cheaper channel exists, evolution uses it instead. It does not buy teaching because it can. It buys teaching when it’s cornered.
That’s the answer to our question. Capacity does not summon the behavior. Cost does. A creature can be a long-lived genius and be taught nothing by its parents, because everything it needed either fit in the genome or came free from watching. And a dim little meerkat teaches, because for one scorpion-shaped problem, nothing cheaper was available.
The child who copies nonsense
So the honest version of the human paradox is not “animals don’t teach and we do.” An ant teaches. It’s stranger than that — and here we finally meet the chimpanzee.
The classic nonhuman cases are strikingly domain-specific. The tandem-running ant teaches one route to one food source. The meerkat teaches one prey-handling trick — dead prey, then disabled, then live, as the pup matures. The pied babbler teaches its chicks to link one “purr” call with food. Each looks less like a school than a single-purpose appliance, installed because that one skill had no cheaper route. Human teaching is the opposite: general-purpose. We teach prey handling, mathematics, manners, music, metaphysics, and the correct way to load a dishwasher — sometimes all before lunch.
But teaching, on its own, still isn’t the deepest trick. Here is where the chimpanzee earns its place in the title. Put great apes and human children through the same broad battery of cognitive tests — and Herrmann and colleagues did, with chimps, orangutans, and two-and-a-half-year-old children — and the children were not simply better at everything. On the physical world, understanding objects and space and quantity, the toddlers and the apes came out about even. The children’s advantage was concentrated in social cognition — the skills for learning from and understanding other minds. (Note the boundary: this is toddlers versus apes on one battery. It does not show that adult humans barely out-think a chimp about the physical world. It shows where the human edge already lives at two and a half.)
Which is, at last, the hedge I promised at the top. “The smartest animal that ever lived” is doing quiet work, because our advantage is not what the phrase implies. It is not a bigger general-purpose processor grinding out every problem better; on raw physical-world cognition a two-year-old and an ape are startlingly close. Our edge is specialized. We come with in-built wiring an ape essentially lacks — above all a hunger to read other minds and a brain pre-built for language, neither of which is cultural, both of which are as much a part of our biology as an eye. And that specialized wiring is precisely what lets us do the thing the rest of this essay is about: it is the socket into which the external inheritance plugs. Not a better engine. A different one — one built, from the factory, to be finished by other people.
The sharpest demonstration is a puzzle box. Show a chimp how to open it, with some steps that are visibly pointless — tap here, wave there, then slide the latch. When the box is transparent and the causal structure is visible, the chimp drops the pointless steps and just gets the food. When the box is opaque and the causes are hidden, the chimp copies the pointless steps too. The human children copied the full sequence either way — transparent or opaque, causes visible or not.
In the transparent case, the chimp is the better engineer. It sees the pointless flourish and says, in effect: cute ritual — I’ll take the snack. The child looks dumber; it copies the flourish anyway. But that “dumbness” is the whole trick, and the difference is strategic, not a matter of ability. Chimps can imitate — in the opaque case they copy the useless steps too. What they are far less inclined to do, especially when the causes are right there in front of them, is preserve another individual’s apparently pointless action. The child does. And the child behaves as though the unexplained step might carry information it can’t yet see — perhaps it makes the box work, or perhaps it is simply how we open boxes. That second reading matters as much as the first. It is the same disposition that later absorbs a table manner, a ritual, a rule of grammar, a line of law: copy the form now, even the parts whose reason is invisible, on the bet that the reason exists — causal or social — and will surface later. High-fidelity copying of other people, faithful enough to carry the parts you don’t understand, is the thing humans do and chimpanzees mostly don’t.
If you think copying the steps you don’t understand is a childish failure we grow out of, I have some news about the most sophisticated manufacturing on the planet. Intel builds chips by a doctrine it named — capital letters, underline, exclamation point, they were not subtle — Copy EXACTLY! When Intel proves out a new process in its development fab and then builds a high-volume factory to run it, the mandate is to replicate the original down to the finest detail: the same equipment, suppliers, plumbing, clean-room conditions, and training methods. And the copying goes absurdly deep — engineers have duplicated variables as fine as lighting, air humidity, and even the maker and color of the paint on the walls, because paint can outgas contaminants that affect the silicon. Not because anyone is superstitious about wall color. Because modern chip fabrication is so complex that nobody can fully model which variables matter in advance, and the era when engineers “improved” the process on transfer was the era of mysterious yield collapses and ruined ramps. So Intel banned improvement. Copy everything, including the parts you’re sure are irrelevant, because you cannot reliably tell the load-bearing step from the ceremonial one — and the cost of guessing wrong is a dead fab. That is the puzzle-box child, in a clean room, wearing a bunny suit. It is over-imitation promoted to corporate law, by some of the most capable engineers alive, precisely because it works.
Banning improvement sounds like the opposite of engineering, but it is the correct response to a system too complex to predict. When you cannot compute in advance which variable is load-bearing, the safe move is to copy the entire known-good answer and let reality — not your intuition — be the editor. Fidelity is one of the ways a system stores answers it cannot yet derive. Which is roughly how DNA works too: replicate faithfully, don’t “improve” the working genes, let selection be the editor over generations. Faithful copying is how you keep an answer whose reasons you don’t have.
The memory outside the skull
Why does faithful copying matter so much? Because it is the only way to preserve a piece of knowledge whose reason nobody currently has — and knowledge whose reason no single person holds is exactly the kind that accumulates. This is the mechanism anthropologists call the ratchet. One generation does something; the next copies it faithfully, including the parts it doesn’t understand; occasionally someone adds an improvement; and the next generation copies the improved version, so it never slips back to zero. High-fidelity transmission is the pawl that stops the ratchet unwinding — the piece without which improvements leak away faster than they arrive. It is not the only ingredient — language, teaching, cooperation, sheer population size, and the division of cognitive labor all feed the ratchet too, and cultures can climb by more than rote copying. But fidelity is the load-bearing one, because everything else is wasted if each generation has to rediscover the base. Turn the ratchet enough times and you get a canoe, a cathedral, calculus, the common law, anesthesia, the iPhone, and the hard-won knowledge that one should never assemble IKEA furniture while hungry — none of it invented by any single mind, all of it the compounded work of thousands who never held the whole.
That is what humans did — not “we teach,” because an ant teaches, but that we combined high-fidelity imitation, open-ended teaching, and language into a channel that no longer stores knowledge inside any one head. We store it between us — in words, writing, tools, institutions, the culture itself — where it ratchets across generations and outlives everyone who fed it. Other animals have the beginnings of this; recent work shows chimpanzees socially acquiring skills no individual invented, so the line is not culture-versus-no-culture. It is that we did it at a scale, a fidelity, and an open-endedness nothing else on Earth has come near — and we spend the first two decades of every human life loading a child into the result.
That’s the difference between you and the chimpanzee. Not mainly the brain behind your eyes. The memory outside it — and the odd, faithful, copy-the-parts-you-don’t-yet-understand disposition by which you were plugged into it before you were old enough to ask why.
What the crib was telling you
Now the paradox reads correctly.
Humans did not trade instinct for intelligence. That is the part the flattering theory got backwards. The genome did not stop building the parenting system; you still arrive wired for the job in all the ways that can be wired in — pulled toward infant faces and cries, primed by hormones and touch, equipped to attend, attach, imitate, and care. What evolution did was subtler and stranger. It built a developmental program whose completed form requires a social world — one that expects a crucial part of the answer to arrive from other people, and is therefore unfinished until it does. A human infant is not born speaking Hindi or Hungarian, but nobody calls language “uninstinctive.” It is a structured program that runs on social input. Parenting is the same. What looks like missing instinct is really instinct built to be completed from outside.
And that is exactly why we need the grandmother, the midwife, the class, and — a thousand years ago, and still across most of the world — the whole village. Not as sentiment, and not as emergency tech support for a defective animal. The village is part of the animal’s expected environment — the place where a species that stores knowledge between its members keeps the half of the parenting program a wasp keeps entirely in its genes. Cut a human off from that world and mothering can fail, which is close to what we see in a great ape raised without other apes. The wasp never needed the connection. We are built to assume it.
So, no — you were not born knowing what to do with your baby. An animal like you couldn’t be, because so much of what would finish you was stored outside you, waiting. Your helplessness at the crib is not the opposite of your intelligence. It is intelligence’s peculiar developmental cost — the price of being the animal whose other half lives beyond the skull.
Are you smarter than a chimpanzee? Not because the brain behind your eyes is a bigger general-purpose engine — on the raw problems it barely is. Because that brain came wired to read other minds and to hold language, and those two gifts plug you into an inheritance no chimpanzee has at anything like your scale: dead people can teach you, strangers can correct you, discoveries can pile up for centuries, and you will faithfully preserve a step whose purpose nobody alive can yet explain.
The wasp had it easier. The wasp also cannot read this.
If you liked this, you’ll enjoy my book, The Science of Free Will — which asks, among other things, why we can’t trade with ants (they can detect cancer), what that has to do with the future of AI, and why a deterministic universe still needs a Supreme Court. Available on Amazon or anywhere else books are sold.
The pre-frontal cortex has much to answer for! Where are the cetaceans on this parenting scale?
Awesome essay, Samir! I sent it to me son and his wife, who are new parents.