Dunbar's Number: The Limit Is Attention, Not Memory
The anthropologist Robin Dunbar noticed a correlation across primate species: the size of the neocortex relative to the rest of the brain predicted the typical size of a social group.
Extrapolating to humans gave a figure around 150 — and when he looked for that number in human societies, he reported finding it repeatedly. In the typical size of traditional villages, in the size of military units that function without extensive formal hierarchy, in the size of communities that hold together on personal relationships alone.
Dunbar's number is now widely cited as the limit on how many stable relationships a person can maintain. It is a genuinely interesting idea, and it is also frequently over-stated, so both parts are worth covering.
Layers rather than a single limit
The most common misuse is treating 150 as a hard cap on people you can know. That is not the claim.
The figure refers to relationships you actively maintain — people whose situation you roughly know, who you would not feel awkward approaching, and where there is some ongoing mutual obligation. Recognition is different: people recognise thousands of faces without difficulty.
Dunbar described nested layers, each roughly three times the size of the one inside it. A very small innermost circle of perhaps five people you are closest to. Around fifteen close friends. Around fifty friends. Around 150 stable relationships. Beyond that, larger circles of acquaintances and recognisable faces.
The layering is arguably more useful than the headline number, because it captures something people recognise: attention is finite, so a larger inner circle necessarily means less depth in each relationship. Adding someone close usually means someone else drifts outward, not because of any decision but because the hours do not expand.
The proposed constraint is maintenance cost. Relationships decay without contact. Keeping one current takes time and attention, both strictly limited, so the total number is capped by arithmetic.
Where the number seems to show up
Several observations are commonly cited as supporting it, and they are worth reporting alongside the caveats.
Group sizes in organisations. Units below roughly this size often coordinate through personal familiarity, while larger ones tend to develop formal structure — written processes, explicit roles, management layers. Some organisations have deliberately split units on reaching a threshold, on the reasoning that informal coordination stops working beyond it.
Community cohesion. Groups that rely on everyone knowing everyone — small settlements, close-knit organisations — appear to function differently past a certain size, requiring rules where familiarity previously sufficed.
Social platforms. Although people accumulate far larger contact lists, studies of actual interaction frequently find a much smaller set of people who are genuinely engaged with. The list grows; the active circle does not, by nearly as much.
That last point is the most practically interesting. Technology has vastly increased how many people you can reach and appears to have changed the number you actually maintain much less — which suggests the constraint is attention rather than access.
How firmly to hold it
Honesty requires flagging that this is a contested finding rather than an established constant.
The original extrapolation was from a small dataset. Correlating brain measurements with group sizes across a limited number of primate species, then projecting to humans, involves substantial assumptions. Later analyses using different methods and larger samples have produced confidence intervals so wide as to make a specific number close to meaningless — some ranging from tens to many hundreds.
Human social organisation varies enormously. People sustain very different network sizes depending on culture, occupation, and circumstance. A single figure across all of that is implausible on its face.
The supporting examples are selected. Group sizes in human institutions vary widely, and it is not difficult to find counterexamples. When a number is famous, confirming cases get noticed and non-conforming ones do not — which is confirmation bias operating on the evidence base itself.
So the defensible version is weaker than the popular one, and still worth having: there is a real limit on how many relationships anyone can actively maintain, it is set by finite time and attention rather than by memory, and it is smaller than the number of people most of us are nominally connected to.
The practical implications survive the uncertainty about the exact figure. Coordination that runs on personal familiarity breaks down as groups grow, and the fix is explicit structure rather than more effort. Contact lists are not relationships. And adding a close relationship costs attention that has to come from somewhere — which is a useful thing to know even without a precise number attached to it.
