Hill & Dunbar (2003)

Citation: Hill, R. A., & Dunbar, R. I. M. (2003). “Social network size in humans.” Human Nature.

Hill and Dunbar (2003) test and refine the Dunbar‑number hypothesis by examining empirical evidence on human social networks. Their goal is to determine whether there are consistent limits on the number of relationships people can actively maintain, and whether those limits show the layered structure predicted by earlier theory.

The paper reviews multiple datasets on human social relationships and finds that social networks are layered, with characteristic sizes at roughly 5, 15, 50, and 150 relationships. The innermost layer consists of a very small number of intimate ties (family or closest friends), and each successive layer includes relationships that are less intense but still meaningful. This structure appears robust across contexts.

A key finding is that active social engagement is constrained by time and cognitive resources, not by the number of people one could theoretically contact. People can “know” many more than 150 individuals, but they cannot sustain stable, reciprocal relationships with more than about that number. This is consistent with the idea that relational maintenance requires effort—regular interaction, mutual awareness, and updating of trust and history.

Hill and Dunbar also emphasize that these limits are not merely cultural artifacts. They appear across societies and remain stable despite large differences in technology and communication tools. This supports the view that the constraint is rooted in human cognition and time budgets rather than purely in social norms or infrastructure.

The paper’s contribution is empirical grounding. It strengthens the argument that relationship capacity is a binding constraint and that any system reliant on trust and repeated interactions must contend with this limit. It also clarifies that the Dunbar number is not a rigid ceiling but a statistical tendency: some individuals maintain slightly larger or smaller networks, but the overall distribution clusters around the same layers.

In practical terms, the research suggests that relationship‑based coordination does not scale indefinitely. Organizations that rely on personal trust, referral networks, or dense interpersonal ties will face a natural upper bound unless they build systems that reduce the cognitive costs of relationship maintenance.

The empirical evidence confirms layered social networks with a stable upper bound on meaningful relationships. Cognitive and time constraints shape social and economic structures in predictable ways—no amount of technology has yet changed the underlying numbers.