Dunbar's Number: The 150 Problem

You have 847 Facebook friends. You follow 1,200 people on Twitter. Your LinkedIn network claims 500+ connections. Your company has 10,000 employees.

But how many people do you actually know?

Robin Dunbar, a British anthropologist, proposed an answer in the 1990s: about 150. That's the number of stable social relationships a human can maintain. It's determined not by choice but by brain size—specifically, by the size of the neocortex.

Dunbar's Number has become one of the most cited findings in social science. It appears in business books, organizational theory, network science, and countless think pieces about social media. It's been used to design military units, structure companies, and explain why Facebook feels empty.

But what does it actually mean? And what happens when we try to exceed it?

The Primate Data

Dunbar's insight came from studying primates. Different primate species live in groups of different sizes. Chimpanzees form bands of about 50. Gorillas live in groups of 10-20. Baboons aggregate in troops of 50-100.

What determines group size? Dunbar found a correlation: the size of a primate's neocortex predicts its average group size. Larger brains (specifically, larger neocortices) enable larger social groups.

The relationship is remarkably tight. If you plot neocortex ratio against group size across primate species, you get a clear line. The bigger the thinking part of the brain, the more social relationships the animal can track.

Why would brain size limit group size? Because maintaining relationships requires cognitive work. You need to remember who's who, who's allied with whom, who owes what to whom, who's trustworthy and who isn't. This is computationally expensive. The bigger the group, the more tracking required.

Social grooming is the mechanism. Primates maintain relationships through grooming—picking parasites and debris from each other's fur. It's time-intensive. As group sizes increase, the time required for grooming increases exponentially. At some point, you literally can't groom enough to maintain all your relationships.

Humans solved part of this problem with language. Talking is more efficient than grooming—you can "groom" multiple people simultaneously through conversation. This allowed our groups to grow larger than other primates'. But there's still a limit.

Language extended our social range but didn't eliminate the constraint. You can talk to a crowd, but you can't have relationships with a crowd. The bottleneck shifted from grooming time to cognitive tracking—and the neocortex is still the limiting factor.

The 150 Prediction

Dunbar extrapolated from the primate data to humans. Given human neocortex size, the predicted group size is about 150.

Then he checked. Is there evidence that human groups cluster around 150?

Hunter-gatherer bands average about 150-200 members at the "clan" level—the largest stable grouping that considers itself a distinct unit.

Neolithic villages from archaeological evidence tend to max out around 150-200 before fissioning into separate communities.

Military units across cultures and eras converge on similar sizes. The Roman century was nominally 100 men (actually 80). The modern company is typically 100-200 soldiers. The Swedish tax authority's research on optimal organizational size landed on 150.

Hutterite colonies (an Anabaptist communal group) deliberately split when they reach about 150 members. They've learned over centuries that larger communities develop coordination problems.

Gore-Tex (W.L. Gore & Associates) famously structured itself around units of about 150. When a factory approached that size, they built a new one. The company's culture of egalitarian cooperation seemed to break down above that threshold.

The number keeps appearing. Not because 150 is magic, but because something about human social cognition creates similar constraints across different contexts.

The precision is misleading. It's not exactly 150—it's probably somewhere between 100 and 250, varying between individuals and circumstances. But the order of magnitude is consistent. Humans can maintain meaningful relationships with roughly 150 people, give or take. Not 15. Not 1,500.

The Layers of Intimacy

Dunbar's work refined the picture. The 150 isn't a flat network—it has structure.

5: Your closest support group. The people you would call in a crisis at 3 AM. Your inner circle.

15: Your sympathy group. Close friends you see regularly and care deeply about.

50: Your regular social contacts. Friends you'd invite to a party.

150: Your "clan." People you know as individuals, whose relationships you track.

500: Your acquaintance network. People you recognize and can place.

1,500: People whose names you can match to faces.

Each layer is roughly three times larger than the one before it. And each layer involves less emotional closeness, less time investment, and less reliable cooperation.

The structure implies trade-offs. You can have more acquaintances, but only by having fewer close friends. Your 150 slots are somewhat flexible, but the total cognitive budget is fixed. Adding a new close friend often means someone else drifts to a more distant layer.

This is why social media feels hollow. You might have 847 Facebook friends, but you don't have 847 relationships. You have the same ~150 relationships you always had, plus hundreds of weak ties that don't function like relationships at all.

The layers also suggest why some organizations feel warmer than others. A startup with 12 people operates mostly in the "sympathy group" layer—everyone knows everyone well. A company with 50 operates in the "regular contact" layer—friendly, but not intimate. Once you cross 150, you're dealing with people you barely know as individuals. The feel changes completely.

What 150 Means for Institutions

If humans can only maintain about 150 genuine relationships, what happens when institutions exceed that size?

Hierarchy. The primary solution is to create structure. You don't need to know everyone in a 10,000-person organization—you need to know your team and your boss, who knows their team and their boss, and so on. Hierarchy is a technology for coordinating beyond Dunbar's Number.

Roles instead of persons. Large institutions relate through roles, not individuals. You interact with "the HR department" or "the vendor" rather than with people you know personally. This is more efficient but less trustworthy.

Formal systems. Contracts, procedures, rules, metrics—these substitute for the personal knowledge that enables cooperation in small groups. You can't rely on everyone knowing everyone's reputation, so you create systems that (try to) enforce good behavior impersonally.

Subcultures. Large organizations develop distinct subgroups—departments, teams, shifts—that function as quasi-independent units. Each subgroup can be Dunbar-sized even if the whole organization isn't.

The transition from personal to impersonal coordination is never smooth. There's a zone between ~150 and ~500 where organizations are too large for everyone to know everyone but too small to have fully developed formal systems. This is often where dysfunction concentrates.

Many rapidly growing startups hit this zone hard. At 50 people, coordination was easy—everyone knew the culture, the priorities, the informal norms. At 300 people, the organization is too big for that but hasn't developed the formal structures to replace it. The result is chaos, politics, and declining effectiveness. Companies that survive this transition often describe it as the hardest period of their growth.

The Trust Problem

Dunbar's Number is ultimately about trust.

In a group of 150, you can track reputations. You know who keeps their promises and who doesn't. You know who's skilled at what. You know the web of relationships—who's married to whom, who owes what to whom, who has feuds with whom. This knowledge enables cooperation without formal enforcement.

This is why gossip matters. In small groups, gossip is the reputation system. It's how you learn who's trustworthy without having to test everyone yourself. It's how the community enforces norms without police. Gossip gets a bad reputation (ironically), but it's essential social infrastructure.

Beyond 150, you can't track all that. You need substitutes for personal knowledge: credentials that signal competence, contracts that enforce promises, institutions that vouch for members, reputation systems that aggregate ratings.

These substitutes work, but imperfectly. They're gameable. They lose information. They create perverse incentives. The transition from personal trust to institutional trust is always lossy.

Consider credentials. A degree from a reputable university is supposed to signal competence. But the degree measures something different from what employers actually need. The credential substitutes for personal knowledge of the candidate—and the substitution is imperfect. Everyone knows people with fancy credentials who are incompetent, and competent people without credentials. But in large anonymous labor markets, we need something.

This is why small communities often have higher trust than large ones. It's not that small-town people are morally superior. It's that reputation works in small groups. Bad actors can't hide. Defection is punished by social exclusion. The incentives favor cooperation.

In large anonymous societies, reputation doesn't work the same way. You can cheat and move on. You can hide behind institutions. The same person who would never steal from a neighbor might happily defraud a faceless corporation. The incentives have changed because the social environment has changed.

This helps explain why corruption tends to increase with organizational size. In a small organization, everyone knows what everyone else is doing. Corrupt behavior is visible and costly. In large organizations, opacity creates opportunities. No one person knows the whole picture. Misconduct can be hidden in complexity.

The institutional response—more oversight, more auditing, more compliance—is a costly substitute for the cheap coordination that small groups achieve automatically. Every dollar spent on compliance is a dollar spent compensating for the loss of Dunbar-scale coordination.

The Digital Dunbar

What does social media do to Dunbar's Number?

It doesn't increase it. Your brain can still only maintain about 150 genuine relationships. The technology doesn't change the cognitive constraint.

But it changes the composition. You can now maintain weak ties across distance that would previously have atrophied. The person you knew in college, the cousin across the country, the former colleague—social media keeps these connections alive (sort of) with minimal maintenance.

And it creates a new category: the parasocial relationship. You might "know" a podcaster or influencer quite well—their opinions, their life events, their personality. But they don't know you. This is one-directional knowledge that feels like relationship but isn't.

The result is often worse, not better. Social media creates an illusion of rich social networks while actual close relationships atrophy. Time spent scrolling feeds is time not spent maintaining the 5 or 15 that actually matter. The feeling of connection is hollow.

Dunbar himself has studied this. His research suggests that online interactions don't substitute for in-person contact—they supplement it, at best. And when online interactions crowd out real-world relationships, wellbeing declines.

Implications for Institutions

Understanding Dunbar's Number suggests several principles:

Small is beautiful—for coordination. Groups under 150 have coordination advantages that larger groups lack. If possible, structure work so that day-to-day coordination happens within Dunbar-sized units.

Hierarchy is expensive but sometimes necessary. The bureaucratic apparatus required to coordinate beyond 150 is overhead. It has costs. Sometimes those costs are worth paying. Often they're not.

Trust scales badly. Personal reputation works in small groups. Larger groups need institutional substitutes—and those substitutes are imperfect. Don't expect large-group cooperation to work as smoothly as small-group cooperation.

Technology doesn't change the constraint. Digital tools can help manage the logistics of large organizations, but they don't increase how many relationships your brain can track. Techno-optimism about bypassing Dunbar's Number is usually wrong.

The Takeaway

Dunbar's Number is the cognitive limit on stable social relationships: about 150 people. It appears consistently across human societies, from hunter-gatherers to corporations, because it reflects constraints of the human brain.

Beyond 150, different coordination mechanisms kick in—hierarchy, roles, formal systems. These work, but they're not as rich or reliable as personal knowledge.

Every large institution faces the 150 problem. The question is how to manage the transition from personal to impersonal coordination while losing as little trust and effectiveness as possible.

The answer, usually, is nested structure—Dunbar-sized subgroups within larger organizations. But this is imperfect. Scale always costs something.

Further Reading

- Dunbar, R. I. M. (2010). How Many Friends Does One Person Need? Faber & Faber. - Dunbar, R. I. M. (1998). "The Social Brain Hypothesis." Evolutionary Anthropology. - Hill, R. A., & Dunbar, R. I. M. (2003). "Social Network Size in Humans." Human Nature.

This is Part 5 of the Anthropology of Institutions series. Next: "Göbekli Tepe"