Dunbar's Number: Why You Cap Out Near 150

Dunbar's number is a proposed cognitive limit of roughly 150 stable relationships you can actively maintain at once. Coined by Oxford anthropologist Robin Dunbar in the early 1990s, it emerged from a correlation between primate neocortex size and social group size, then extrapolated to the average human brain. It is a soft average, not a hard wall.

Where 150 Actually Comes From

The figure is not a guess. Dunbar's 1992 paper, "Neocortex size as a constraint on group size in primates," found that across monkeys and apes, the bigger a species' neocortex relative to the rest of its brain, the larger its typical social group. A tamarin, with a low neocortex ratio, lives in groups of about five; a macaque, with a higher ratio, manages around forty. The neocortex appears to be the organ that tracks who is who and who relates to whom.

Plugging the average human neocortex ratio into a regression built from 38 primate genera produced a predicted group size of about 148, which Dunbar rounded to 150. He was candid that this was exploratory: the 95% confidence interval ran from roughly 100 to 230, and the human neocortex ratio sits well outside the primate range the regression was built on, so the extrapolation is admittedly a stretch. He then checked the prediction against the real world and found echoes of it in hunter-gatherer bands, Neolithic farming villages, English settlements in the Domesday Book, the basic Roman military unit, and the size of people's Christmas-card lists.

The deeper claim is the "social brain hypothesis": primates did not grow large brains to use tools or find food, but to manage the political complexity of group life. Tracking alliances, debts, grudges, and reputations is computationally expensive, and the neocortex is the hardware that pays for it.

The Nested Layers: 5, 15, 50, 150, 500, 1500

The 150 figure is really the middle of a fractal series. Each circle of intimacy is about three times larger than the one inside it, the circles are cumulative (each contains everyone closer in), and emotional closeness falls off as they widen. You can estimate your own version of these layers with the Dunbar's number calculator.

• ~5 (support clique): the people you would turn to in genuine crisis. You contact them at least weekly and give them roughly 40% of your total social time.
• ~15 (sympathy group): close friends whose death would cause deep grief. Contacted at least monthly, and together with the inner five they soak up most of your social attention.
• ~50 (good friends / affinity group): people you would invite to a large dinner; the bond is real but lighter.
• ~150 (active network): everyone you have a genuine reciprocal relationship with and contact at least yearly.
• ~500 (acquaintances): people you know but do not actively maintain.
• ~1500 (recognizable faces): the outer ceiling, roughly the number of faces you can attach a name to.

These layers are not static. Relationships form a "convoy" that travels with you through life, with people drifting inward and outward as circumstances change. Studies of email, mobile-phone logs, Facebook, and Twitter have all recovered roughly the same layered scaling, which is part of why the model has held the public imagination so firmly.

Grooming, Gossip, and the 42% Problem

Dunbar's original puzzle was why monkeys spend so much time picking through each other's fur. Grooming, he realized, is social glue, and the time a species spends on it scales with group size. Run the numbers forward for a human group of 150 and the math turns absurd: maintaining those bonds by primate-style grooming would demand around 42% of waking time. Chimpanzees, our closest comparison, spend only roughly 15% to 20% on grooming.

His resolution is elegant. Language is "grooming at a distance," and several times more efficient because you can talk to a small group at once (Dunbar estimates conversation clusters average about three listeners) while your hands stay free. Gossip, in this view, is not a vice but the core technology that let human groups grow past the size other apes can sustain. Even so, the time cost never disappears. Friendships, unlike kin bonds, carry a measurable decay: stop meeting face to face for a few months and a friend begins sliding outward into a mere acquaintance, which is why the layers require constant reinvestment to hold their shape.

Real-World Uses, and the Live Debate

The most cited application is W.L. Gore & Associates, maker of Gore-Tex. Founder Bill Gore noticed that once a plant passed a certain size, "we decided" quietly became "they decided," so he capped each factory (commonly reported at around 150 people) and built a new one next door rather than expanding. Strikingly, Gore died in 1986, years before Dunbar published, having found the limit by trial and error. The Swedish Tax Authority reorganized offices around a similar ceiling, and military companies have hovered near it for centuries.

The science is genuinely contested. A 2021 study in the Royal Society journal Biology Letters by Patrik Lindenfors and colleagues reran Dunbar's logic with modern phylogenetic statistics. Their Bayesian method produced central estimates of roughly 69 to 109 people, while a generalized-least-squares method produced roughly 16 to 42, and both came with enormous 95% confidence intervals (about 4 to 520 and 2 to 336 respectively). They argued that no precise number is statistically defensible. The honest takeaway is that the headline figure is a fuzzy average around a real phenomenon: human social capacity is finite, layered, and time-bound, even if "150" is a convenient round number rather than a law of nature. If patterns and scaling laws interest you, the Zipf's law calculator explores a similar rule in language and city sizes.

What survives the debate is the underlying insight. Your attention is a budget, not an infinite resource, and the structure of how you spend it (a tiny core, a slightly larger circle of confidants, and a wide ring of casual ties) shows up again and again across cultures and even across digital platforms. The number you land on matters less than the shape.

Frequently Asked Questions