Software development teams are social systems, just like those formed by other primates. So, what can we learn from these primates?
The way humans form social organizations impacts our industry, raising questions about the ideal size for a team or company. Dunbar’s number, often cited as 150, is commonly referenced as the size of a group humans fell comfortable with. However, this interpretation is simply incorrect. Dunbar’s original scientific publication says something entirely different.
Dunbar studied non-human primates. He found a correlation between the volume of their neocortex - a part of the brain - and the size of the groups they form. From this data, Dunbar extrapolated a group size of 147.8 for humans. As with all scientific findings, there’s a margin of error: with 95% probability, the group size lies between 100.2 and 231.1. Another paper, however, suggested a completely different range, with a confidence intervals from 3.8 to 292.0. This research also presents some other data analyses, all of which offer confidence intervals ranging from single digits to several hundred. Based on this, the Dunbar’s number has no practical value.
Language instead of Social Grooming
But Dunbar’s primary focus wasn’t the number. In groups, primates engage in social grooming not only to remove parasites but also to strengthen social bonds. Dunbar highlighted the relationship between the time spent with social grooming and the size of the primate group. Larger groups need more social grooming. For the size of human groups, the time needed for social grooming would be unfeasible, leading Dunbar to suggest that humans developed language as a more efficient way to maintain social cohesion. Thus, in his opinion language didn’t evolve for hunting or tool-making, as other scientists have proposed, but to nurture social relationships.
In other words, Dunbar’s central thesis is not the size of the group but that human language evolved to strengthen social bonds.
Dunbar’s number of 150 is just the extrapolated maximum size for human groups that use language as a process analogous to social grooming. He mentions clans or villages in this context. He also describes other groups: “bands” with 30 to 50 members and “tribes” or “subtribes” with 1,000 to 2,000 people.
Thus, Dunbar’s paper discusses a variety of human group sizes - not just one. Groups significantly different from 150 simply belong to another category. His thesis isn’t about the group size per se but about the mechanisms a group uses to sustain itself—and he clearly states this.
One interesting point: Dunbar sees military groups as supporting his thesis, noting that the military has groups of 100 to 200 members. However, there are also much smaller military units, like a squad (2 to 8 soldiers in the German army), or larger ones, like a battalion (300 to 1,200 soldiers), which he doesn’t examine further.
There is extensive criticism of Dunbar’s paper from other scientists, so virtually every part of has been debated. For example, the group size findings are challenged by evidence of fission-fusion groups among primates. In these groups, individuals join and leave, sometimes sleeping together in one place but spending the day apart. Such groups are temporary, and species with this behavior require less social grooming time, often forming larger groups. So actually primates can form large groups without complex human language.
Why Is Dunbar’s Number So Interesting?
For organizing human teams, Dunbar’s number is not helpful. Dunbar himself states that human groups can be practically of any size. You don’t even need to read the extensive critiques to reach this conclusion.
The critiques also raise other interesting points. For example, it’s unclear why the number of people a person knows would affect group size. Even if we only know a limited number of people and speak to them regularly, a group can still be much larger. It’s enough for people to act in coordinated ways. Thanks to language, humans can coordinate on a large scale, up to nations and beyond. It’s obvious that we have different levels of familiarity and trust with different people. We leverage this in daily work: instead of giving information directly to someone, we might ask a third person to do it because they have a better relationship with that person.
Reason for the Misinterpretation
To me, the misinterpretation of the Dunbar’s number reflects a deeper problem: human and social behavior is oversimplified. In the end, we get a number that supposedly represents the ideal group size — an easy rule to follow.
But intuition should tell us otherwise. Everyone knows from their own experience that people can operate in groups of different sizes — in both private and professional contexts: the company, clubs, the neighborhood, friends, family. These groups vary in size. For particularly large groups, there are hierarchies, like in the military, but also in companies, with teams, departments, and offices.
These groups often don’t last long. For example, during a training or the first consulting meeting, trainers and consultants work with a group they’ve never met before — and it works. This is obviously a different group dynamic than with one’s family, but such groups have different goals.
You could argue that trust grows over time. But trust can also form quickly: when a patient is admitted to the hospital, they may entrust their life to the treating doctor, even though they’ve never met them before.
Certainly, Dunbar’s research can serve as inspiration for thinking about mechanisms that strengthen group cohesion. His thesis is that language evolved to strengthen social bonds, and he provides statistics on how much time people spend talking about social relationships and gossip. Measures to strengthen team cohesion, like informal conversations, can be useful. Where does your team have such a forum? It doesn’t have to be a team-building events; something like a regular shared lunch can serve the same purpose.
tl;dr
Dunbar’s number doesn’t tell us anything about the ideal size for teams or companies. They can be any size and structured in different ways. The misinterpretation of the number suggests that our industry is susceptible to oversimplifications that even contradict intuition. Teams need a mechanism to foster social cohesion.
This is a translation of my German blog post at heise Developer.