Our brains grew
to hold each other.

Some animals can live alone. Tigers do. Humans cannot. Other people are not something we happen to enjoy. They are something we run on, the way we run on food and sleep.

That sounds like a nice thing to say. It is actually a finding, and a very well-tested one. Four things we now know:

• People with weak social ties die sooner. Not by a little. The researchers who measured it put the risk alongside the ones your doctor already warns you about, smoking included.
• Being left out switches on the same part of the brain as being physically hurt. A painkiller dulls it.
• Ten hours with nobody makes your brain crave people the way ten hours without food makes it crave a meal.
• And our brains seem to have grown large in the first place to keep track of each other.

This ledger gathers that evidence in one place — from biology, medicine, brain scans, archaeology, history — and follows it all the way through. Why we are social, what being social built, and what we have done to it in the last hundred years.

Psychology has a bad habit. A charming study goes viral, and years later nobody can make it happen again. So this ledger keeps to four rules, written down here so you can catch it breaking them:

• Big, or repeated. Studies that pooled dozens of other studies, or followed tens of thousands of people, or have been redone by other labs and still worked. If a claim rests on one clever experiment, the card says so.
• Agreement across fields. The findings trusted most here are the ones a doctor, an ape researcher and a brain scientist reached separately.
• The source, always on screen. Every card shows who found it and where, at the bottom, linked. Go and check.
• Doubt out loud. Where the evidence is shaky, argued over, or resting on one lab, it is flagged in this colour. A whole part near the end is nothing but what we still do not know.

No finding here is tidied up to sound better than it is. It turns out the honest version is usually the more interesting one anyway.

In 1965 a public-health team asked 6,928 adults in one Californian county some ordinary questions. Are you married. Do you see friends. Any relatives nearby. Do you belong to a church, a club. Then they put the answers in a drawer.

Nine years later they checked who had died. The people with the fewest ties were dying at 2.3 times the rate of the most connected men, and 2.8 times the rate among women.

The obvious complaint is that this is backwards — sick people stop going out, so being alone is a symptom, not a cause. The researchers had the numbers to check. The gap stayed after they accounted for how healthy people said they were at the start, and for income, smoking, drinking, weight, exercise and doctor visits.

It was the first time someone measured loneliness the way you measure cholesterol. It behaved like cholesterol.

One study can be a fluke. So a team collected every good long-term study they could find — 148 of them, following 308,849 people — and pooled the lot.

The answer: having strong relationships came with a 50% better chance of being alive at follow-up. It held for young and old, men and women, healthy and already ill, whatever they died of, however long the study ran. When something shows up that consistently, it is usually not an accident of who was in the room.

One detail is the real giveaway. When researchers measured connection crudely — do you live alone, yes or no — they found about a 19% better chance. When they measured it properly, asking how woven into other lives a person really was, it rose to 91%.

Better measuring found more of it. That is what happens when a thing is really there.

There are two ways to be lonely, and they are easy to mix up. One is a fact about your diary: few people, rarely. The other is a feeling: you can have a full week and still feel unseen. Which one does the damage?

A study pooling 70 studies and 3.4 million people separated them. Both mattered, by about the same amount. Being cut off raised the risk of dying early by 29%. Feeling lonely raised it by 26%. Living alone, by 32%. The feeling was not a watered-down version of the fact. It was its own risk.

And one result almost everyone gets backwards: the effect was stronger in people under 65, not older. Loneliness is filed in most minds as a problem of old age. As a health risk, it is worse for the young.

The setup was almost too simple. Lying in a scanner, people played a computer game of catch with two others they thought were real. Halfway through, the other two quietly stopped throwing them the ball.

Nothing was at stake. No money, no faces, no conversation, no reason to ever meet these people again. And a patch of brain called the dorsal anterior cingulate cortex got busier. That patch is known for something specific: it handles the part of physical pain that bothers you. The more left out someone said they felt, the more it lit up.

We did not pick the word "hurt" for this by accident. It is a description of the wiring.

If hurt feelings really do run on the pain system, then a pain pill ought to take the edge off them. That is such a strange prediction that it makes a good test. There is no obvious way for it to come out right by luck.

It came out right. Sixty-two people took either paracetamol (what Americans call Tylenol) or a sugar pill every day, not knowing which, and each evening wrote down how much their feelings had been hurt. Over three weeks, hurt feelings dropped in the pill group and stayed flat in the sugar-pill group. In a second study, people on the drug showed less activity in those same pain areas while being rejected in the lab.

Now the honest part. These were small groups, and a later independent study did not get the brain result again. So this is not medical advice, and one study is not proof. It is a strange prediction that came true, which is the kind of thing worth noticing.

This is the neatest experiment in the whole field, because it does the thing you would do if you could: take the same people and starve them of two different things.

Forty adults spent ten hours completely alone. On another day, the same forty spent ten hours without eating. Then they went into a scanner and looked at pictures — some of food, some of people together.

A small cluster of cells deep in the middle of the brain, the ones that produce the feeling of wanting something, answered both times. After the hungry day they fired at the food. After the lonely day they fired at the people. And the stronger someone said the craving felt, the stronger the signal was.

Same machinery, aimed at whatever is missing. Loneliness is not feeling sad about people. It is being hungry for them.

Think about how thirst works. It is horrible on purpose. An animal that felt only mildly bothered about water would wander off and die. The awfulness is the whole point — it is what makes you get up and do something.

The standard explanation of loneliness, worked out by John Cacioppo, is that it belongs to that family. For an early human, being away from the group was not a mood. It was no one to watch while you slept, no help if you broke a leg, no share of the food. A feeling that switched on when you were alone and refused to switch off until you got back would be worth every bit of its unpleasantness.

Two things fit that story. Roughly half of the difference in how lonely people feel is inherited — 48% in a study of 8,387 twins — which is what you expect of built-in equipment rather than a personal failing. And similar states turn up in other social animals, not just us.

There is a sting in it, though. The alarm does not only push you toward people. It also makes you watch them for signs of danger — sensible if you are genuinely exposed, ruinous if you are just lonely, and the reason it can trap the person it is trying to rescue. That trap gets its own section later.

The alarm is working perfectly. What changed is that we built lives where it can ring for years with nothing to answer it.

Every act of kindness is a small loss. You give away food you could have eaten, time you could have used, risk you did not have to take. Over millions of years, that should lose.

And mostly it does. Most living things are not kind. The puzzle is not why animals are selfish — that needs no explaining — but why a few are not, and why one ape became so extravagantly unselfish that it will donate blood to a stranger it will never meet.

Biology has found, in the end, only a handful of ways this can happen. This part walks through them in order, and then looks at what they built when they arrived in an animal like us. The answers are less romantic than you would like and more beautiful than you expect.

That was J. B. S. Haldane, reportedly in a pub, and it is not a joke about how much he liked his family. It is arithmetic.

You share about half your genes with a brother, and about an eighth with a cousin. So from a gene's point of view, saving two brothers or eight cousins breaks even with saving yourself. William Hamilton turned that into a formula in 1964, and it is now one of the load-bearing walls of biology: help others when the benefit to them, discounted by how related you are, beats the cost to you.

It explains a great deal that would otherwise look mad. Why bees die for the hive. Why a squirrel screams a warning that gives away its own position. Why almost every human society, however different in every other way, treats family as a special case.

What it does not explain is the interesting part. A friend shares no more genes with you than a stranger does. On this theory alone, friendship should not exist at all.

Here is an oddity you have never thought to find odd. Almost no animal keeps living long after it stops reproducing. Human women routinely live thirty or forty years past menopause. From evolution's cold viewpoint, that is a lot of upkeep for no offspring.

Kristen Hawkes and colleagues spent years recording who actually gathers the food among the Hadza of Tanzania. The answer was startling: older women were among the hardest workers, digging deep tubers that small children cannot get at themselves, and handing the food to their just-weaned grandchildren.

That does something specific. It lets a mother stop feeding her toddler and have her next baby sooner, without the toddler suffering. A grandmother has no more children of her own, but she gets more grandchildren — so genes for living long and helping could spread.

Honestly held: this is built mostly on observation rather than experiment, and other explanations are still argued for. But it reframes old age completely. Not the leftover part of a life. Possibly the reason our lives are long at all.

Robert Trivers asked how helping a non-relative could ever pay, and found that it can — but only under conditions. Three of them:

• You meet again. If this is the only time you will ever see each other, there is no later in which to be repaid.
• You recognise and remember each other. A favour is worthless if the other one cannot tell you apart from anyone else, or forgets by morning.
• Cheats get caught. If you can take without giving and nobody notices, taking wins, and the whole arrangement collapses.

Meet those three and something remarkable becomes possible: I help you now, you help me later, and we both end up better off than either of us managed alone. Trivers argued this is where a whole shelf of human feelings comes from — gratitude, trust, suspicion, guilt, the hot flare of anger at being used. They are the accounting system.

Hold on to the three conditions. Much later in this ledger, when we get to cities and screens, you will watch all three of them come apart.

This is the part people skip, and it is the part that does the work. A group of generous people with no way of spotting a taker is not a stable group. It is lunch.

So wherever biology has built cooperation between non-relatives, it has also built bookkeeping. Cleaner fish that nibble a client honestly get invited back; ones that bite get chased off. Vampire bats share blood with bats who shared with them. And humans developed the most elaborate cheat-detection equipment in nature: we remember who owes what for years, we feel a physical jolt at unfairness, and we talk about each other constantly.

Which recasts something you were probably taught was a vice. Gossip is not a character flaw in our species. It is the audit trail that lets kindness survive.

You cannot interview the Stone Age. But you can carefully study people still living by hunting and gathering, and the Hadza of Tanzania are among the most studied on earth.

Researchers asked 205 Hadza adults two things: who would you want in your camp next season, and — handing over three straws of honey — who would you give these to. From the answers they drew the network.

It looked eerily familiar. A few people with many ties and most with few. Friends of your friends being friends. Ties returned rather than one-sided. Closeness falling off with distance. People clustering with others who give at the same rate they do — the generous with the generous, the tight-fisted with the tight-fisted — and that similarity reaching two steps out through the network.

The shape of a human social world is not a modern invention. We brought it with us.

A chimpanzee mother carries her infant almost constantly, and will not let another adult hold it for months. Among humans, a baby is passed to a grandmother, an aunt, an older sister, a neighbour, sometimes within the first day.

Sarah Hrdy's argument is that this single difference is the root of much of what we are. Human infants are extraordinarily expensive and slow, and a mother alone could not have afforded them. Across many cultures, roughly half of the care a child gets after early infancy comes from someone who is not its mother. We are, in the technical sense, a cooperative breeder — and no other ape is.

The consequence is a chain reaction. A baby that must be cared for by several people has to be able to charm several people — to hold a gaze, read a mood, work out who is willing. Hrdy's claim is that our talent for reading each other did not appear so we could hunt or scheme. It appeared because a human baby had to win over a room.

An adult a child has never met is hanging washing and drops a clothes-peg. He reaches for it, fails, and reaches again. He does not ask for help. He does not look at the child. He just struggles.

Eighteen-month-olds toddle over and pick it up for him. Not for a reward — none is given — and not because anyone asked. Across a set of different situations, children this young worked out what the adult was trying to do and helped him do it. Young chimpanzees did it too, in a simpler way, on some of the tasks.

It is worth pausing on the age. This is before real sentences, before school, before anyone has been told to share. Two things have to already be running for it to happen: the child has to understand what someone else wants, and has to want them to get it.

We do not learn to care about each other. We arrive furnished.

Brains are ruinously expensive. Yours runs constantly, cannot be switched off, and burns a fifth of everything you eat. Evolution does not hand out organs like that for nothing.

So what was the bill for? The old answer was tools and food — that we got clever to crack open nuts and outwit prey. The evidence has not been kind to it. The thing that best tracks brain size across the primates is not diet, or tool use, or the difficulty of the landscape.

It is how many others you live with.

Robin Dunbar lined up primate species and plotted two things against each other: the size of the neocortex — the folded outer layer that does the thinking — and the typical size of the group that species lives in.

They rose together. Tamarins live in groups of about five. Macaques live in groups of about forty, and have proportionally much more neocortex. Run the same line out to a human brain and it predicts a group of about 150 — which turns out to be roughly the size of a great many human things, from Neolithic villages to army companies to the number of people you send cards to.

The logic is that relationships are computationally brutal. With five companions you track a handful of connections. With fifty you track hundreds, and you also need to know who is annoyed with whom, who owes whom, who is lying. Company is the hardest sum a primate ever has to do.

Now the dissent, because it is real. A 2021 reanalysis argued the statistics do not hold up, that diet and range predict brain size better, and that the confident figure of 150 has a confidence interval wide enough to drive a bus through. Dunbar has answered at length. Treat the direction as well supported and the exact number as folklore.

This is the cleanest test anyone has run on why humans are different, and it works by refusing to flatter us. Give chimpanzees, orangutans and two-and-a-half-year-old children the exact same battery of puzzles, before the children can read or have been to school. Then compare the scorecards.

On the physical problems — where is the food hidden, which pile is bigger, how do you get a treat out of a tube, what happens if I tilt this — the toddlers and the chimpanzees came out about level. Human toddlers were not better at physics. At all.

On the social problems the gap was enormous. Learning by copying somebody. Following a pointed finger. Working out what another creature was trying to do. There, the children left both ape species far behind.

We are not the animal with the better processor. We are the animal that can read the others.

Point at a thing across the room, and a human baby will look where you pointed. That seems like nothing. It is not nothing.

To follow a point, the baby has to work out that you have something in mind, that you want them to have it in mind too, and that the finger is a signal rather than an object. Chimpanzees, raised by humans, mostly do not get there. They will follow your gaze to find food for themselves. What they largely do not do is point things out to each other for no reason except that the other might want to know.

Michael Tomasello's name for what humans have and apes lack is shared intentionality: not just wanting a thing, but holding a goal jointly, knowing the other holds it too, and knowing they know you do.

It is contested. Some researchers argue apes do more of this than Tomasello allows. But the size of the human version is not really in dispute. Everything we build — a language, a bridge, a law — rests on two people being able to want the same thing on purpose.

Wood, graphite, clay, lacquer, glue, the metal band, the rubber. The mining, the milling, the kilns. Nobody alive knows the whole process for a pencil, and nobody ever did.

Joseph Henrich's argument is that this is the answer to the entire riddle of human success. Our advantage was never the individual brain — take a modern human, drop them alone in a wilderness their ancestors thrived in, and they will very probably die. Our advantage is that we can store know-how in the group and pass it on faithfully, so each generation starts where the last one stopped.

That makes intelligence a property of a network, not a skull. And it makes a testable prediction, which is where this stops being philosophy: a bigger, better-connected population should invent more and lose less. Modelling and the archaeological record both point that way.

You are not smarter than a hunter-gatherer. You are plugged into more people.

At the end of the last ice age the water rose and the Bass Strait flooded, turning Tasmania from a peninsula into an island and stranding a few thousand people.

Over the following millennia the Tasmanian toolkit got simpler. Bone tools, cold-weather clothing, fishing equipment — things the archaeology shows they had once used — thinned out and in some cases disappeared. Nobody got stupider. The population was simply too small, and too cut off, to keep every skill alive. Each craft needs enough practitioners that the knowledge survives an unlucky death, a bad decade, a generation that does not take to it.

This is the collective-brain claim running in reverse, and it is the part that should worry you. If know-how lives in the connections rather than in the heads, then connection is not a nice-to-have. It is the storage medium. Cut it, and things drain away that no one decided to abandon.

The Tasmanian case is argued over — how much was lost, and how much was sensible adaptation to a changed island. The principle it illustrates is much better established than this one example.

Researchers have done the unglamorous work of eavesdropping — sitting in cafés, canteens and trains, sampling what people genuinely talk about rather than what they say they talk about.

The result is consistent and slightly embarrassing. Around two thirds of conversation time goes on social topics: who did what, who is with whom, who behaved badly, personal likes and dislikes, other people's business. Politics, sport, culture, work, the weather — no single one of them reliably gets even a tenth.

Dunbar's reading of that is that we have language for a purpose we would not have guessed. Monkeys and apes hold groups together by grooming, hands in fur, one partner at a time. That does not scale — you cannot groom 150 individuals. Talking does scale: you can bond with three people at once, and you can find out about people who are not even in the room.

Gossip is not language's decline. On this account it is language's job, and grammar is what it needed to do it.

Put someone in a brain scanner, give them a hard puzzle, and certain regions light up. Then take the puzzle away and tell them to rest. Most of the brain quiets down. A particular set of regions does the opposite: it gets busier.

That set has a name — the default mode network — and for years nobody was sure what it was for. It looked like idling.

Then people noticed something. The map of the resting brain and the map of the brain thinking about other minds were largely the same map.

Working out what is in another person's head is one of the hardest things a brain does. It uses a specific set of regions: the middle of the front of the brain, a patch behind and above each ear, and a region at the back of the midline.

Those are also, to a large degree, the regions that come up when you are asked to do nothing. Pooling many imaging studies, and then tracing which regions are actually wired to which, the two networks overlap substantially — though not perfectly, and the overlap is more exact in some parts than others.

The most careful reading is not "the brain's default is social". It is that social thinking is what the brain falls into when it is left alone, and it is left alone a lot.

Worth knowing: the default mode network is not only social. Remembering your own past and imagining your own future run through much of it too. What those have in common with reading other people is an open question, and a good one. You appear to think about yourself with the same equipment you use to think about everybody else.

Married women were put in a brain scanner and shown a signal meaning an electric shock might come. Sometimes they lay there alone. Sometimes they held a stranger's hand. Sometimes they held their husband's.

Alone, the threat lit up the regions that handle alarm and bodily arousal. Holding a stranger's hand, that response was smaller. Holding their husband's hand, it was smaller still.

And the size of the calming tracked how good the marriage was. Women who reported a better relationship showed less alarm while being held.

Be careful with this one. Sixteen people, one laboratory, and the marriage part is a correlation rather than proof of cause. What makes it worth printing is that the same shape keeps turning up in later, larger work. A hand is not a metaphor. Your alarm system counts it.

The test could not be plainer: crouch against a wall as if sitting on an invisible chair, and hold it until you cannot. Then, separately, count the people in your social circle.

The two go together. People who lasted longer had larger networks.

That sounds like nonsense until you know what the wall-sit is measuring. Enduring discomfort leans on the body's own opioid system — the internal supply of the chemicals that opiate drugs imitate. And that same system is heavily involved in social bonding across the primates. A person with a livelier opioid system should be better at both.

So the test is not really about toughness. It is a rough, indirect reading of the machinery you bond with.

Honestly: it is a correlation in about a hundred people, and pain endurance is affected by many things besides opioids. It is a clue, not a proof.

Students were taught simple dance moves and put in groups. Some groups moved in time with each other. Others did the same amount of moving out of time. Effort was varied too, so the two could be told apart.

Afterwards, two things were measured: how much pressure they could stand on the arm, and how close they felt to the group. Both synchrony and exertion raised pain threshold, and independently of each other. Being in time also made people feel closer.

Then came the test that makes it more than a curiosity. The dance was repeated while some people were given a drug that blocks the body's opioid receptors. In that group the pain-threshold rise did not happen.

Which is the cleanest evidence we have that dancing together, a thing humans do everywhere and cannot explain, is the body dosing itself.

If the opioid system is really how we bond, blocking it should make people feel lonelier. Someone gathered every study that had tried this: seventeen results, 455 people, all given a drug that blocks opioid receptors, all asked how connected they felt.

The answer is yes, and small. Feelings of connection went down by roughly one point or less on a seven-point scale. Real, consistent in direction, and modest.

The people who did the pooling are blunt about what that means: the studies were too small individually, the drug is not precise, and the true effect may be smaller still. Their own conclusion is that this chemistry is not required — people fully blocked still feel close to the people they love.

So the opioid system turns the volume on connection. It is not the speaker.

This is what an honest finding looks like: a direction you can trust, a size you should not exaggerate, and authors who say so themselves.

You have heard oxytocin called the love hormone, the cuddle chemical, the trust molecule. The idea came from good places: in other mammals it genuinely matters for bonding, and an early human study found that a nasal spray of it made people hand over more money in a trust game.

That result has not replicated well. Reviewing the three ways people have tried to link oxytocin to trust — nasal sprays, levels measured in blood, variants in the receptor gene — the honest summary is that there is no reliable evidence that human trust rides on oxytocin.

None of that means oxytocin does nothing. It means the simple headline was wrong, and it spread anyway, because love hormone is a much better story than complex peptide with context-dependent effects.

Keep this card. It is the reason every other card here names its study and its size.

This is the finding that stops the argument about whether loneliness is "just" a feeling.

Blood was taken from 141 older adults, repeatedly, over time, and researchers read which genes their white blood cells were actually running. In the people who felt more isolated, one pattern showed up again and again: inflammation genes turned up, and antiviral genes turned down.

The pattern has a name — the conserved transcriptional response to adversity — and it is what a body does when it expects physical danger. Wounds need inflammation. Viruses are a problem for later.

Which is a sensible setting for an animal alone on a plain, and a terrible one for a person alone in a flat for twenty years. Inflammation running hot for decades is the common road to heart disease, and to a good deal else.

A pattern in a blood sample is still only a pattern. The same work went one step further, in rhesus macaques, where isolation can be studied rather than only observed.

The monkeys showed the same shift as the lonely humans: the same inflammatory cells multiplying, the same antiviral genes turned down, and the body no longer listening properly to its own stress hormone — the brake that is supposed to shut inflammation off.

Then they were exposed to a virus. The isolated animals handled it worse. The virus grew faster in them.

So this is not a feeling with a biological echo. It is a setting on your immune system, and the setting has consequences when something real arrives.

The honest note: the human half of this is correlation, and the causal half is monkeys. Together they are far stronger than either alone, which is exactly why the study did both.