Dunbar's Number

Transfers

Robin Dunbar’s 1992 observation that primate neocortex size correlates with social group size, extrapolated to humans, predicts a cognitive limit of roughly 150 stable social relationships. The number has become one of the most influential cross-domain mappings from evolutionary biology to organizational design.

Key structural parallels:

• The cognitive container — each person maintains a mental model of their social world: who is reliable, who is skilled at what, who is allied with whom, who owes favors to whom. This model consumes cognitive resources proportional to the number of relationships tracked. At roughly 150, the container is full. Adding the 151st relationship either displaces an existing one or degrades the quality of all models. This is why people in large organizations report feeling “anonymous” — they have exceeded the capacity of anyone’s social-cognitive container.
• The governance transition — below 150, groups can coordinate through mutual knowledge. Every member knows every other member well enough to predict behavior, enforce norms through reputation, and resolve conflicts through personal negotiation. Above 150, these mechanisms break down. The group must introduce formal governance: rules, roles, hierarchies, processes. This transition is not a management choice but a cognitive necessity. Organizations that resist formalization above 150 do not stay informal; they become chaotic.
• Nested circles — Dunbar’s later work refined the single number into a series of concentric circles with characteristic sizes: 5 (intimate support group), 15 (close friends), 50 (good friends), 150 (meaningful contacts), 500 (acquaintances), 1500 (faces you can name). Each circle represents a different depth of social knowledge and a different coordination mechanism. Software teams (5-7 people), departments (15-50), and divisions (150) map onto these circles with striking regularity, suggesting that organizational architecture converges on cognitive constraints regardless of industry.

Limits

• False precision — Dunbar’s own research gives a range of 100-250, not a precise 150. The number has been reified through repetition into an engineering constant when it is actually a statistical central tendency with wide variance across individuals and cultures. Using 150 as a hard threshold for organizational design decisions (splitting teams, capping company size) imports a specificity that the data does not support.
• Communication technology as cognitive prosthetic — the original primate data is about grooming-based bonding, which requires co-presence and physical contact. Humans already transcended this constraint with language. Digital communication tools (social media profiles, CRM systems, shared documents) further externalize relationship maintenance, functioning as cognitive prosthetics that may shift the ceiling. LinkedIn users “maintain” hundreds of relationships that would have been impossible to sustain through face-to-face interaction alone. Whether these tool-mediated relationships are “real” in the sense Dunbar’s model requires is an open question, but dismissing them entirely ignores the mechanism that distinguished humans from other primates in the first place.
• The correlation is disputed — several replication attempts have questioned the neocortex-to-group-size correlation, and a 2021 study by Lindenfors, Wartel, and Lind concluded that the confidence interval for the human prediction is so wide (2-520) as to be uninformative. Dunbar’s Number may be a real phenomenon observed across human groups that happens to lack the neat neuroscientific explanation originally proposed for it.
• Cultural variation — the converging evidence (Neolithic villages, military units, Hutterite communities) comes predominantly from Western and small-scale societies. Whether the same number applies to cultures with fundamentally different kinship structures, collectivist orientations, or social organization patterns is less established. The number may reflect a specific mode of relationship maintenance rather than a universal biological constant.

Expressions

• “We’ve hit Dunbar’s Number” — used when an organization experiences coordination breakdown at roughly 150 people, diagnosing the problem as structural rather than managerial
• “Keep it under Dunbar” — used in organizational design discussions to argue for capping team, department, or company size
• “The Dunbar layer” — used to describe the social-cognitive tier at which a group transitions from informal to formal coordination
• “Two-pizza teams are Dunbar’s inner circle” — connecting Amazon’s team-sizing heuristic to Dunbar’s 5-person intimate support group

Origin Story

Robin Dunbar, a British anthropologist and evolutionary psychologist, published the foundational paper “Neocortex Size as a Constraint on Group Size in Primates” in 1992. The paper correlated neocortex ratio (neocortex volume relative to total brain volume) with mean social group size across primate species, then extrapolated the regression line to the human neocortex ratio, yielding a predicted group size of roughly 150. Dunbar subsequently found converging evidence in human social structures: Neolithic farming villages (150-200 inhabitants), Roman military centuries (80-160 soldiers), Hutterite community split threshold (150), Gore-Tex factory unit cap (150), and Christmas card distribution lists (mean 153.5 recipients). The number entered mainstream discourse through Malcolm Gladwell’s The Tipping Point (2000) and has since become a standard reference in organizational design, software team scaling, and social network analysis.

References

• Dunbar, Robin I. M. “Neocortex Size as a Constraint on Group Size in Primates” (Journal of Human Evolution, 1992) — the original paper
• Dunbar, Robin. How Many Friends Does One Person Need? (2010) — popular-audience treatment of the number and its nested circles
• Gladwell, Malcolm. The Tipping Point (2000) — popularization of Dunbar’s Number for business audiences
• Lindenfors, Patrik, Andreas Wartel, and Johan Lind. “‘Dunbar’s Number’ Deconstructed” (Biology Letters, 2021) — critical reanalysis questioning the precision of the prediction
• Kerr, Dave. “Hacker Laws” — https://github.com/dwmkerr/hacker-laws