PsychologyCounterintuitive resultDebated
Dunbar's Number
Also known as Social Brain Hypothesis, Cognitive Social Limit
You can maintain real relationships with about 150 people. Beyond that, a group becomes something different. Is this a law of human nature?
- When
- 1992
- Difficulty
- Intermediate
- Reading time
- 7 min
Counterintuitive result. Nothing is wrong with the reasoning. The result is simply correct and human intuition is simply unreliable here. Not a contradiction at all.
Discover
In 1992, Robin Dunbar plotted neocortex size against social group size across 38 primate species. The correlation was strong. Baboons with larger neocortices live in larger stable groups. Chimpanzees, with even larger neocortices, live in groups of around 50. The relationship held consistently across the primate order.
Then Dunbar did the extrapolation for humans, whose neocortices are the largest of all: the predicted social group size was approximately 148.
He then looked at the ethnographic record. Hunter-gatherer bands across cultures and continents averaged around 100–200 people. Neolithic farming villages, reconstructed from archaeological data, clustered around 150. Hutterite religious communities — which Dunbar found particularly interesting — consistently split when they reached about 150, because their members reported that the community had become too large to manage without formal rules.
The military had independently discovered the same number: Roman legions were divided into 80-man centuries; modern military companies range from 100 to 250 people. The functional unit — the group within which every member personally knows every other member's capabilities and can be relied upon without formal command structures — consistently hovers around 150.
Understand
Dunbar's Number is the hypothesis that the human brain can maintain stable social relationships — relationships involving knowledge of who each person is, their status, and their relationship with others — with approximately 150 people.
This is not a claim about acquaintances (you can recognise thousands of faces) or connections (you may have thousands of LinkedIn contacts). It is a claim about cognitively active relationships: relationships that require maintaining a mental model of the person, their history, and their social position.
The neocortex — the most recently evolved part of the brain — is disproportionately enlarged in social primates. Dunbar's Social Brain Hypothesis proposes that the neocortex evolved primarily to manage complex social relationships, not for toolmaking, language, or abstract reasoning (though these came with it). The cognitive demands of tracking many social relationships constrain group size more fundamentally than food supply or physical territory.
The layered model
Dunbar and colleagues subsequently found that human social networks are not a flat structure of 150 equal relationships. They are layered:
- 5: intimate circle — family, closest friends, people you speak to most frequently
- 15: sympathy group — close friends and family, people whose death would deeply distress you
- 50: band — friends, people you see regularly
- 150: clan — Dunbar's Number — the full stable community
- 500: extended community — acquaintances you recognise and greet
- 1,500: outer edge — people you can recognise by name
Each layer is roughly three times the size of the previous one, and requires progressively less cognitive and emotional investment.
The shape of it
Same shape, different costume:
Examples
| Community type | Typical size |
|---|---|
| Hunter-gatherer band (ethnographic average) | 100–200 |
| Neolithic village (archaeological reconstruction) | ~150 |
| Hutterite community before fission | ~150 |
| Roman century (legionary tactical unit) | 80 |
| Modern military company | 100–250 |
| Gore-Tex factory (deliberate Dunbar limit) | ~150 |
| Average Christmas card list (UK studies) | ~150 |
| Maximum Facebook "close friends" actively tracked | ~150–200 |
W.L. Gore & Associates — makers of Gore-Tex — famously limits each factory to about 150 employees, having found that larger factories required management structures that smaller ones did not. This was not based on Dunbar's research, but the parallel is often cited.
Explore
Does social media change the number? The most common challenge to Dunbar's Number in contemporary discussion is: does Facebook or…
Does social media change the number? The most common challenge to Dunbar's Number in contemporary discussion is: does Facebook or Instagram allow us to maintain more stable relationships than our neocortex could manage before? Studies of Facebook data have found that, despite having hundreds or thousands of "friends," most people actively engage with only about 150–200 connections — and the core group they interact with is much smaller. The cognitive limit appears to persist even when technology expands the number of contacts nominally maintained.
Critiques and challenges. Several researchers have challenged Dunbar's Number directly. A 2021 paper by Patrik Lindenfors and colleagues, re-analysing the primate data with different statistical methods, found no significant correlation and argued the number was an artefact of the analysis. Dunbar disputed the methodology. The debate continues, and the specific number of 150 has a wider confidence interval than popular citation suggests — somewhere between 100 and 250 in most ethnographic estimates.
Language as the social grooming upgrade. Dunbar proposes that language evolved partly as a scalable alternative to physical grooming. Primates maintain social bonds through grooming (physically removing parasites), which is effective but time-consuming — limiting group size to how many individuals you can groom. Language allows "social grooming" at larger scale, as conversation allows bonding with multiple people simultaneously. This is why, Dunbar argues, human conversation is predominantly gossip — tracking the social relationships of others is the primary ancestral function of language.
Institutional design. The Dunbar hypothesis has been applied to organisation design. Companies and military units that exceed ~150 members tend to develop formalised management structures — job titles, performance review systems, written policies — that smaller groups manage without. This threshold is sometimes interpreted as the point at which social cohesion breaks down without formal institutional support.
Historical context
1992. Dunbar publishes "Neocortex size as a constraint on group size in primates" in the Journal of Human Evolution, establishing…
1992. Dunbar publishes "Neocortex size as a constraint on group size in primates" in the Journal of Human Evolution, establishing the correlation and the predicted human limit of ~148.
1993. Dunbar publishes further work on the Social Brain Hypothesis, connecting the neocortex correlation to the evolution of language.
1998. Dunbar validates the 150 figure against hunter-gatherer ethnographic data, Domesday Book villages, and military unit sizes in a widely cited paper.
2000. Malcolm Gladwell's The Tipping Point discusses Dunbar's Number in the context of how communities maintain coherent social norms, giving it wide popular circulation.
2010s. Social media provides large datasets for testing the number in digital environments. Multiple studies confirm that active engagement networks remain clustered around 150–200, despite theoretically unlimited connectivity.
2021. Lindenfors et al. challenge the original statistical analysis. Dunbar's team responds. The debate about the specific mechanisms and the robustness of the 150 figure continues.
Why it matters
Organisation design. If human groups cannot maintain stable informal social cohesion beyond 150, then organisations exceeding…
Organisation design. If human groups cannot maintain stable informal social cohesion beyond ~150, then organisations exceeding this size need explicit management infrastructure — not as an efficiency choice, but as a cognitive necessity. Many startups that struggle with their cultural shift as they grow from 50 to 200 employees may be experiencing Dunbar's Number in practice.
Community formation. Digital platforms that aim to create community — forums, Discord servers, online groups — face Dunbar's constraint: communities that grow past 150–200 members without structural intervention (subgroups, moderation hierarchies) tend to lose the social cohesion of the original small community. Knowing this threshold allows community designers to plan structural interventions before they become necessary.
Social capital in cities. Humans evolved in small groups but increasingly live in cities of millions. Urban social life is characterised by networks of small communities (neighbourhoods, workplaces, friend groups) each roughly Dunbar-sized, embedded in larger anonymous structures. The tension between evolved group psychology and urban anonymity shapes much of urban sociology.
Modern relevance
Remote work and team cohesion. Remote work eliminates many of the casual social interactions through which team members maintain…
Remote work and team cohesion. Remote work eliminates many of the casual social interactions through which team members maintain their cognitive models of each other. Teams that previously maintained natural Dunbar-level cohesion through physical proximity need explicit alternatives. The Dunbar framework provides a vocabulary for what is being lost and why it matters.
AI and social simulation. AI systems are beginning to model social relationships in ways that previously required human cognitive overhead — tracking who knows whom, how relationships are evolving, what each person's status and history is. Whether AI can extend effective human social capacity beyond the Dunbar limit is an open question with significant implications for organisations and governance.
Platform governance. Online communities that understand Dunbar's Number build their moderation and community structures around the 150 threshold. Subreddits, Discord servers, and neighbourhood groups that succeed at scale often do so by maintaining Dunbar-level subgroups within larger umbrella structures.
Sources & further reading
Primary source
(1992)Neocortex size as a constraint on group size in primatesJournal of Human Evolution, 22(6), pp. 469–493.
The paper establishing the neocortex-to-group-size correlation and extrapolating to predict the human social group limit of approximately 150.
How this connects
- Related to
- Bystander EffectBoth concern the limits of what human cognition and social psychology can manage in groups. The bystander effect is about responsibility diffusion; Dunbar's Number is about the cognitive limit on stable relationships.
- Ringelmann EffectBoth suggest that group size has inherent cognitive and social limits. Ringelmann shows effort declining with group size; Dunbar shows that stable relationships require cognitive resources that cap at around 150.
- Matthew EffectThe Matthew Effect operates through social networks — and Dunbar's Number puts a hard ceiling on the stable social networks within which cumulative advantage can operate person-to-person.