The standard story of human cognitive evolution — "bigger brains evolved to make better tools" — is wrong, or at least incomplete. The dominant pressure on primate brain size across the last 30 million years has been social complexity, not tool use. The fossil and comparative evidence (Robin Dunbar's "Social Brain Hypothesis"): the size of the neocortex in any primate species scales with the size of its typical social group, not with its tool complexity. A chimp living in a group of 50 has a smaller neocortex than a human living in a group of 150; a lemur living in a group of 10-12 has a smaller one still. Brains are for managing relationships, not for operating on the physical world. The fact that we can also operate on the physical world (tool use, language, abstract reason) is a side effect.
The implication: the human brain's social architecture is the load-bearing structure, not a side feature. You can damage the language centers and still form relationships. You cannot damage the relationship-tracking centers and still form a person.
Dunbar's empirical finding (and the layered model that follows from it): humans don't have one social circle, they have layered circles, each nested inside the next, with the layer boundaries set by the cognitive capacity to maintain each level.
| Circle | Approximate size | What you do with them | |---|---|---| | 5 | Intimate support group | People you would call at 3am in a crisis. The Dunbar layer most directly correlated with the 50% mortality reduction (friendship). | | 15 | Close friends / sympathizers | People you see regularly and confide in. The actual friends, by most operational definitions. | | 50 | Friends | People you have lunch with, remember birthdays for, and would feel comfortable calling out of the blue. | | 150 | Meaningful contacts | Everyone you would recognize, greet by name, and know your relationship to. The canonical "Dunbar's number" from a 1993 paper. | | 1500 | Recognizables | Everyone whose face you know and you could place a name to, but you don't track your relationship to. | | 5000 | Familiar faces | Everyone you would recognize on the street. The upper bound of human face-recognition capacity. |
Why the layered structure matters for friend-making: the "intimates" tier is the tier that drives the health outcomes (friendship). The "5" is the cap on what you can actually invest in. The 50 / 150 / 1500 circles are receptive — they can move up (a 50-tier person can become a 15-tier person with enough repeated time and disclosure) but they cannot move up without the time cost. Friendship is allocation, not collection. The user cannot out-collect their way to more 5-tier slots; they can only choose which 5-tier slots to invest in.
Dunbar's anchor finding: across all surviving hunter-gatherer societies (which are the closest proxy we have to ancestral human social structure), the typical band size is 100-200, with a hard mode around 150. Beyond 150, the rate of interpersonal violence (homicide, feuds) goes up significantly. This is the operationalization of the "tribe" size. A 150-person band is the largest group in which every member can know every other member by name and by relationship, and where social sanctioning (gossip, ostracism, status games) is sufficient to maintain order. Larger groups require formal institutions (law, police, courts) to manage the conflict that informal sanctioning can no longer police.
The user is 22, mobile, post-college. Their intimate slots are mostly already-occupied by college/high-school friends. Every new friendship they want to make at the intimate level is a reallocation — they have to either deprioritize one of the existing 5 or accept that the new relationship will only get to the 15 or 50 layer.
This is the strategic decision most adults are confused about. The question is not "how do I make more friends" but "given my 5/15/50 budget, which existing or potential relationships am I willing to invest in at which layer?" A 22-year-old with 5 intimate slots and 100+ latent acquaintances has to choose — and the modern friendship ecosystem makes the choice invisible by making all relationships feel equal.
"And go back to Dunbar's number, right? You can't scale. You
cannot scale your friendship network because you were just
cognitively limited."
Followed by Mark: "Friendships don't scale. That's why they're meaningful."
roughly to Dunbar's layers (virtue = 5, pleasure = 15, utility = 50)
substrate that Dunbar's number is a constraint on
synthesis
(2010) — the popular treatment
Language* (1996) — the original social-brain argument
the network-level extensions