Dunbar's Number
An essay in my series on human interdependence
Robin Dunbar is an anthropologist who observed that among primates, larger brain size and larger group size are correlated. He hypothesized that recognizing others as belonging to a group and being able to predict their behavior takes brain power. Extrapolating from data on primates, he predicted that group size among humans would be about 150.
Anthropologists are free to debate the validity of Dunbar’s number for predicting the size of hunter-gatherer bands. What strikes me is that Dunbar’s number predicts the point at which organizations cannot function effectively without bureaucratic structure.
Note that we do not observe any major sports in which the size of a team exceeds Dunbar’s number. Note that we do not observe any K-12 classrooms in which the number of children exceeds Dunbar’s number. And when one finds college lectures with more than the Dunbar’s number in the audience, to the professor the students are a faceless mass. The absence of feedback makes for a generally inferior learning experience.
Think of small organizations as sub-Dunbar, and large organizations as super-Dunbar. My claim is that sub-Dunbar organizations operate informally, while super-Dunbar organizations rely on formal bureaucratic structure.
Super-Dunbar organizations tend to require decisions to be recorded in writing. In ancient times, the invention of writing probably did a great deal to allow groups to expand beyond the Dunbar number. Empires and religions could now be organized as hierarchies. Rules could be codified.
Super-Dunbar organizations tend to have organization charts. When you want a new marketing brochure in a sub-Dunbar organization, you either write it yourself or just “call Jane.” In a super-Dunbar organization, you have to make a request to the marketing department.
Super-Dunbar organizations tend to give new hires written procedure manuals and formal training. They often assess performance using explicit criteria and on a regular schedule.
Software development in sub-Dunbar organizations can be handled by one or two stars, who work quickly and retain important knowledge in their heads. Super-Dunbar organizations are often less agile, and they require much more documentation of their systems.
Formal hierarchies, rules, and procedures are appealing to some people and appalling to others. Some people who perform well in a small organization are dysfunctional in a large institution, and conversely.
An important error to avoid is to expect a large organization to work like a small one, or conversely. “From each according to his abilities, to each according to his needs” is a reasonable approach for relating as a family. I would not want to relate to my family using prices and incentives. But for a large scale society, prices and incentives are far better. It is inconceivable that we could organize the production of the vast array of goods and services that we enjoy by telling people to just go out and use their abilities as best they can and take whatever they think they need.
This essay is part of a series on human interdependence.
Ok, relevant topic: In my experience the 150 number sort of holds up. Perhaps it’s somewhat lower. Maybe 100 or so. But it may vary depending on domain.
The post caused me to re-think what has been a mystery of my work experience: Without going into too much detail, I had a job where I was effective for a period of time, and then suddenly found myself bogged down in politics. I frequently complained that I was spending more time negotiating deals with my internal colleagues than with external counterparties. The group and company were growing tremendously, and this effect probably occurred somewhere around the 100-200 mark. As it turns out, I abandoned the team to go work on a new initiative, made up of about 30 people. That group too quickly grew to about 2000 and became completely unwieldy, ultimately leading to its failure. (This was at a company that likes to think of itself as “the world’s biggest startup”, which really just means that it pays very little attention to developing an effective bureaucracy.)
Amazon famously (apocraphally?) has it’s “no meeting with more people than can be fed by one pizza” rule. That would suggest another rule perhaps around 10 people or so. That rule seems to correlate to my experience as well, and to sports teams generally (11 is a common number for team sports with large organized groups).
I wonder how much these numbers correlate to the size of various divisions in military organizations. And I wonder whether the size of those organizations changes during wartime, when you’d think coordination problems would be different and high-stakes than in peace time?
For me, I wonder now, how many workplace issues that one would tend to attribute to “personality differences” are effectively attributable to organizational structure, and indeed, attributable to a simple variable like average group size.
Informal social control is a crucial private-governance mechanism in groups and orgs of any size, insofar as networks form, norms emerge, gossip circulates, and individuals care about reputation and its consequences.
Student-athletes tell me that sports teams have a bedrock norm in dispute-resolution: "It never turns out well when coach gets involved." Accordingly, they live by informal social control, in which captains play a focal role.
Good managers expect subordinates "to work it out" in day-to-day frictions. Employees have contempt for colleagues who go straight to the top in dispute-resolution.
Cop shows always showcase informal social control in the force, as well as mistrust of complex organization (esp. 'encroachment' by other branches of the force, and 'internal affairs' sub-orgs).
Good CEOs worry about boundaries of the firm (what to outsource), incentives (principal-agent problems) *and* firm culture.
The Dunbar number plays a substantial role in super-Dunbar orgs, via emergent informal social control.