Dunbar's number: where 150 came from and what it doesn't mean

In the early 1990s, evolutionary anthropologist Robin Dunbar correlated the relative size of the neocortex across primate species with the average size of their social groups. He extrapolated the regression to humans and arrived at a predicted group size of roughly 150 (Dunbar, 1992).

The number found a cultural afterlife disproportionate to its empirical foundations. Dunbar's number now appears in books on organizational design, network science, and dating apps. It is used as a brute fact about human cognition. The original research is much narrower than that use suggests.

1. What the original study actually claimed

Dunbar's neocortex-ratio correlation was based on a small sample of primate species, and the human extrapolation was at the edge of the regression line (with substantial uncertainty bands). The "150" number had a 95% confidence interval roughly between 100 and 230, depending on the specific analysis (Dunbar, 1993). The popular version dropped the interval.

The claim is also specifically about a category of social relationships — what Dunbar called "the maximum number of people with whom you can have a genuinely social relationship," which he operationalized as "the kind of relationship where you wouldn't feel embarrassed about joining uninvited for a drink." This is a narrower category than the popular usage implies.

2. The critique

In 2021, Patrik Lindenfors, Andreas Wartel, and Johan Lind published a re-analysis using updated primate data, a larger sample, and modern statistical methods. Their conclusion: the original neocortex-group-size correlation is much weaker than Dunbar's analysis suggested, and the human extrapolation produces confidence intervals so wide (anywhere from 4 to 520 individuals depending on the analysis) that it provides little usable information (Lindenfors et al., 2021).

This is not the same as saying "Dunbar's number is wrong." It is saying "the data don't support a specific number with the precision the popular usage assumes."

3. The empirically supportable claims

What is empirically supportable, separate from the specific 150:

Social network sizes do show consistent structure. Across multiple cultures and methodologies, people's social networks appear to have nested concentric layers — closest contacts (3-5), trusted friends (10-15), wider friend circle (30-50), acquaintances (100-200), wider recognition network (500-1500). The layers are real even if the specific numbers vary (Hill & Dunbar, 2003).

Cognitive constraints on social tracking exist. Maintaining a relationship requires memory of shared history, awareness of the other's current context, ongoing interaction. These have real costs.

Active relationships do correlate with attention budget. Studies of phone call frequency, Christmas-card lists, and Facebook interactions all show that people actively maintain a finite number of close ties, even with unlimited contact-list capacity.

What is not well supported is a precise number — 150 or any other — as a hard cap on human relationships.

4. The implication for organizations

Companies and military units have used Dunbar's 150 as a design constraint — keep work units under 150 people to maintain coherence. The advice may be useful, but its justification is more anecdotal than scientific. The post-2021 view is that there's a number where social coherence degrades, and it's roughly in that range, and it varies by context and culture more than the precise figure suggests.

5. The takeaway

Use "Dunbar's number" as a heuristic, not a law. The general shape of the claim — that social networks have nested layers with diminishing intimacy and finite capacity — is well-supported. The specific 150 figure is much weaker. Most popular invocations conflate the two.