In university I had a core group of four. That wasn’t the number of people I knew. On any given Friday, twenty of us might end up at the same bar, shoulder to shoulder, knowing each other’s names and faces and latest dramas. But by 2 a.m., when the noise had thinned and the party had done whatever it was going to do, it was almost always the same four of us left in somebody’s kitchen, talking the way people only talk when nobody’s listening except the people in the room.

I noticed the pattern more than once. The friends I’d made through class had a core of four, surrounded by a looser circle of a dozen. The friends from sports, the same. The dorm, the same. House parties would widen to sixteen or twenty and then narrow back to a smaller group, almost always around four, when the night got honest. I didn’t have a theory about it then. I just knew which rooms I could say the real thing in.

The four of us didn’t all survive graduation. The pattern held while we had the container around us: the dorm, the same bar on Friday, the same kitchen at 2 a.m. Once we scattered into different cities and first jobs and rent that finally mattered, the container was gone, and some of the thread frayed. A couple of those friendships survived the move and are still where the honest conversations happen. The others drifted in the way everyone’s friendships drift when the structure dissolves. The pattern never guaranteed the people in it, only that while we were inside it, something honest could happen there.

The quiet ceiling

The instinct, most of the time, is to widen. We keep adding to the dinner party until the table won’t hold another chair. We fold new people into the group chat until scrolling it feels like work. We say “the more, the merrier” as if more is the point. And sometimes it is. A birthday, a wedding, a band reunion is allowed to be big. But for the kind of conversation that leaves a trace on your week, wider is usually worse.

Something changes past a certain size, and most of us feel it without naming it. The table splits into two conversations, then three. The quietest person gives up and checks their phone. The person with the most social gravity starts performing a little, because now there’s an audience. Honesty leaves the room without anyone noticing, and what’s left is pleasant and shallow, the kind of evening you enjoy and forget.

The research points at four

There’s a reason this happens, and it’s smaller than most social-capital theories. It’s working memory. In a now-classic field study, the anthropologist Robin Dunbar and his colleagues timed how long conversational groups could stay unified before they fractured into sub-conversations. They watched cafés, pubs, and dinner tables and found the ceiling was four. Above four, groups reliably split. The constraint isn’t cultural. It’s the bandwidth of the human working memory in real-time conversation. You can hold roughly three other minds in active tension with your own, and no more.

Dunbar’s broader work on social networks shows the same shape at larger scales: a layered ring of about five closest people, fifteen good friends, fifty friends you’d invite to a party, 150 you’d recognize by name. Each layer roughly three times the size of the one inside it. The five is where the real thing lives. Everything else radiates out from it.

Dunbar's layered social network

Closest (support)

5 people

Sympathy group

15 people

Affinity group

50 people

Active network

150 people

Robin Dunbar, layered social network model (summarised in The Conversation, 2021)

The striking thing about these numbers is how stable they are across contexts no pre-industrial villager could have coordinated. Christmas card lists settle around 150. Telephone-call logs settle around 150. Neolithic village sizes settle around 150. Military companies have centred on roughly that figure for about as long as militaries have been organised. Inside every one of those outer rings, five people still carry most of the weight. Half of that inner circle tends to turn over every seven years unless someone actively protects it, which is another reason to take the small number seriously. Those five are where the work of adult friendship actually has to happen.

What the crowd is actually doing

It would be too neat, though, to say small is always better. Some of the most meaningful experiences a life contains happen in a crowd. A packed wedding dance floor. A stadium singing the same song in unison. A protest. A funeral. A church. Émile Durkheim called this collective effervescence: the feeling that you and everyone around you are briefly part of a larger body. That experience doesn’t scale down. You can’t reproduce it in a kitchen with three friends. The crowd is the point.

But watch what’s actually happening inside a crowd like that. A concert of ten thousand is really one focus at the top of the room, which is the band, and underneath that focus, three thousand little clusters of two and four and six: the pairs squeezed against the rail, the group of five in the pit, the couple standing in the back, the friends who came together and will go home together. You show up to the show with your people, and the mass of strangers around you is the weather, not the room. When something transcendent happens you turn and look at them, the three or four faces you came with. The crowd is what makes the moment big. The small group is where the moment gets shared.

The honest version is quieter than “small beats big.” Depth lives in the small group even when the spectacle is happening outside it. That’s why the best shows live in your memory beside the names of the people you went with, not the size of the crowd.

The case for the outer rings

Before pressing the small-group argument too hard, it helps to sit with the strongest counter-case. In 1973, the sociologist Mark Granovetter published “The Strength of Weak Ties,” which tracked how people actually found their jobs. His finding was that the useful lead almost never came from a close friend. It came from someone further out: a friend of a friend, an old classmate you’d half-forgotten, the former colleague whose Slack handle you still have. The closest five were too similar to you to know anything you didn’t already know. The outer rings held the new information, the unfamiliar rooms, the role that took your life somewhere else.

Fifty years of research has largely kept the shape of that result intact. The people who introduce you to a new idea, a new job, sometimes a new partner, are rarely your closest three or four. Small circles cost you something: they’re not where those introductions happen. The argument for small groups isn’t a dismissal of the outer ones; the two do different jobs. A group chat of fifty is a different tool answering a different question from a group of five, not a worse version of it. The mistake is expecting the fifty to give you what the five can, or the five to give you what only the fifty will.

The software that forgot about the ceiling

Most of the apps we use to keep up with friends don’t acknowledge this shape at all. Group chats scale to fifty, a hundred, two hundred, and pretend the experience above the line is the same as below it. Instagram flattens a close friend and a distant acquaintance into two scrolls of the same feed. The underlying assumption is that bigger is always more: more reach, more engagement, more value. Past a certain size that stops being true, but the product keeps suggesting another person to add.

It’s part of why we built Colloquies with a cap. Between two and eight people per circle, nothing bigger. That’s a design choice with real cost. It rules out communities, it rules out fanbases, it rules out anything that uses social-graph growth as its engine. We’re making the bet that the cost is worth it because the ceiling is real. Whether eight is exactly the right upper number is something we’re still watching and will probably tune over time. That there needs to be an upper number, we’re sure about.

The room that empties out

There’s an uncomfortable truth inside all of this. If you want to be deeply known by many people, you’ll have to give up on being deeply known at all. The capacity is small. The time is shorter than it feels. Most of the work of adult friendship is choosing who your five or six are going to be, and then showing up for them often enough that the group stays a group instead of a set of individual names drifting apart.

I didn’t choose the room that made me feel most like myself in university. It was just where the four of us ended up at 2 a.m. when everyone else had drifted home, a kitchen with the kettle on and nothing left to prove to anybody. Maybe the question worth sitting with is what would have to narrow in your life for a room like that to exist in it again.

Hero image by Andrew Amelianovich on Unsplash.