Dunbar's Number: How Many Relationships Can You Really Maintain?

Dunbar’s number is the idea that a person can keep up stable, meaningful relationships with about 150 people at a time. The Oxford evolutionary psychologist Robin Dunbar arrived at it in the early 1990s by taking a pattern from primates, whose groups are larger in species with larger brains, and asking what it predicted for humans.¹

So how many relationships can you maintain? Nobody can give you an exact figure. In 2021, three researchers at Stockholm University reran the primate analysis with larger datasets and newer methods, and concluded that the uncertainty was far too wide to support the famous figure, or any other single number, as a limit.² Dunbar replied that they had used the wrong statistics, and that the networks people actually keep still cluster near 150.³ What survives the argument is more useful than the number itself: your time for people is limited, and you spend it very unevenly. In practice, that means deciding which handful of people get most of your hours and which contacts you keep warm with an occasional note.

Prefer to watch? The video covers the same ground as this article.

VideoDunbar's Number: How Many Relationships Can You Really Keep Up?WiserHours

Where Dunbar’s number came from

Dunbar’s number came from a line drawn through primate data. Dunbar’s 1992 work linked group size to the size of the neocortex, the brain’s outer layer, relative to the rest of the brain across primate species, and in 1993 he used that relationship to predict a typical human group size of about 150.³

The reasoning is called the social brain hypothesis. Living in a stable group means coordinating with others and defusing conflicts, and Dunbar argued that keeping track of these relationships takes brainpower, so the size of the neocortex sets an upper limit on the group.²

His 1993 paper added a second limit, time. Monkeys and apes hold their groups together by grooming each other, and grooming a group of human size would make intolerable demands on the day; Dunbar suggested that language evolved as a more efficient way to bond, and that the size of conversation groups fits that idea.⁴

He then checked the prediction against real groups. Hunter-gatherer societies are layered, with families inside bands and bands inside larger communities, and when Dunbar looked up the sizes, the community level came out almost exactly on the predicted figure.¹

Extending a line beyond the data it was drawn from is like reading an adult’s likely height off a growth chart made for children: the trend may be real, but the further you go past the data, the less exact the answer. The human figure is an extrapolation from other species, so it depends heavily on how the line is drawn and which data go into it.

Dunbar’s figure, like the later estimates that challenge it, is a predicted average for a group, as the Stockholm team pointed out.² Nobody’s brain shuts the door on the next new friend, and nobody fails at networking by knowing more than 150 people.

The Christmas-card study and the circles of 5, 15, 50 and 150

The best-known test of Dunbar’s number in everyday life counted holiday cards. In a 2003 study, Russell Hill and Robin Dunbar measured people’s social networks through the Christmas cards they exchanged, and found that the full networks those cards reached averaged close to the predicted 150, and somewhat smaller when counting only the people contacted directly. The paper describes its setting as contemporary Western society, and it found that contact depended both on circumstance, such as distance, and on closeness, which let the authors separate out layers within the network.⁵

The appeal of the measure is easy to see: a card is a small yearly gesture, so a card list gives a rough, everyday record of who someone still keeps in touch with. But Dunbar’s later review lists the sample as a few dozen people.⁶

Those layers became the more durable part of the idea. A 2005 analysis by Wei-Xing Zhou, Dunbar and two colleagues pooled published data on human groups and found sizes that step up by a factor of roughly three: a support clique of a few people you would turn to in a crisis, a sympathy group of about a dozen or more whom you contact at least monthly, then larger bands, and the circle of about 150.⁷

1. About 5: the people you would turn to in a crisis; they get the largest share of your social time
2. About 15: adds the people you contact at least monthly; together with the first circle, they take more than half of social effort
3. About 50: mostly friends, seen less often than the inner circles
4. About 150: your full personal network, where family members predominate; all four sizes are averages from Dunbar and colleagues, and individuals vary

The reason layers form is the same time budget. According to Dunbar’s 2020 review, more than half of people’s social effort, measured as how often or how long they interact, goes to the people in the two innermost circles.⁶ Once most of your hours go to a handful of people, only thin slices remain for everyone else.

Newer data give the layers partial support. In a 2016 study of a year of calls on a European phone network in 2007, Mac Carron, Kaski and Dunbar found a clear layered pattern: the innermost and outermost layers matched earlier estimates, but the layers in between varied widely.⁸

Picture your own circles as a hypothetical. The innermost might hold a partner, a sibling and two close friends. The next adds the people whose news you would want to hear the same week. Most of your work contacts, from the former manager you like to the client you email twice a year, sit in the outer circles, where contact is rare and closeness is modest. Judge your circles by where your time goes.

The 2021 reanalysis: too uncertain for any single number

Not as a precise limit, according to a 2021 paper in Biology Letters by Patrik Lindenfors, Andreas Wartel and Johan Lind of Stockholm University. They rebuilt the primate calculation with more species and two modern statistical approaches, and found that the prediction for humans shifted with every choice of method and data, with ranges too wide to name a number.²

The study

Mixed evidence

Lindenfors, Wartel & Lind (2021), Biology Letters

Across eight models, the predicted average human group size changed with every choice, from the teens to about 110. The model the authors judged the best test predicted about 69 people, with a 95% confidence interval from about 4 to 292, and across all models the upper ends reached 520. The authors concluded that specifying any single number is futile and that a cognitive limit on human group size cannot be derived this way.²

A confidence intervalconfidence interval: A range of values within which the figure a study is estimating, such as an average effect, can reasonably be expected to lie. Its width is about precision: how tightly the study pinned that average down, not how much individual results vary.Full entry in the glossary is the range of values the data can reasonably support. When that range stretches from a dinner party to a large village, the method cannot tell 150 apart from much smaller or much larger numbers, which is the heart of the Stockholm team’s case. It is like a forecast of anything between frost and a heatwave: honest, but no help in deciding what to wear. Their test concerned the brain-based derivation, not direct counts of people’s networks, although the authors, who declared no competing interests, went further and said the concept lacks empirical support, citing studies that found a wide variety of human group sizes. That matters beyond biology, because the number has been used as a design rule: according to the Stockholm team, Sweden’s tax agency restructured its offices to stay within the figure.²

So if a book tells you to cap your contacts at 150, or an organization sizes a department around it, treat the figure as a rule of thumb someone chose, not a limit the science has set.

Dunbar’s reply, and what each side gets right

Robin Dunbar rejected the reanalysis in a 2021 article, arguing that the Stockholm team used a regression method that flattens the slope of the line. With the method he used originally, he wrote, the same data predict about 158 rather than the 71 he calculated for their method.¹

His 2023 paper adds two technical arguments. First, primates fall into distinct grades, groups of species with different baselines; fitted to the grade containing apes and humans alone, his equation predicts about 152, with a much narrower range. Second, he argued that the Stockholm team reported the wider kind of interval, suited to a single group, when the question concerns the average group.³

His strongest point is empirical. His 2020 review lists measured personal networks, from Christmas-card lists and phone records to wedding guest lists, and reports an overall mean near 150 across very different kinds of samples, while noting that network size varies with personality, age and sex.⁶

The Stockholm authors answered him in turn. They argued that the method Dunbar favors over-corrects the slope, that his approach applied to their data gave an estimate of about 290, and that each new way of grouping species yields yet another Dunbar’s number.⁹ Other ways of counting add to the doubt: as the 2016 phone study notes, US studies from 1984 and 2001 that used other estimation methods put the average American’s personal network well above Dunbar’s figure.⁸

Our reading, which neither side states in these words, is that the dispute is mostly about how precisely primate brains can predict a human number, and much less about whether people sort their relationships into a few close and many distant ones. You can therefore drop 150 as a quota and keep the rough idea: the relationships you actively keep up probably number somewhere between several dozen and a few hundred, not the thousands a contacts list can hold, and your personal figure depends on who you are, as the variation in Dunbar’s own review shows.

What Dunbar’s number means for keeping up a professional network

For your working life, the useful part is the budget. In a 2015 study, close friendships faded faster than family ties, and less where people kept putting in effort, so keeping a network alive is a question of where your limited hours go. The evidence comes from Sam Roberts and Robin Dunbar, who followed 25 school-leavers in England for a year and a half as they moved to university or work.¹⁰

Only about half of the friends who started in someone’s inner circle were still there at the end, against more than two-thirds of family members. Friendships declined less when people invested more effort in them: for the young women, mostly by talking more, and for the young men, mostly by doing more things together.¹⁰ It is a small sample at one life transition, so take it as a sketch of how friendships drift.

Time together also tracks how new ties form. In two studies, of adults who had recently moved and of first-year students, Jeffrey Hall found that more hours spent together went with closer friendships and, among first-year students followed over nine weeks, with changes in closeness.¹¹ This is correlational, and people who like each other spend more time together, but it points the same way as the decay study.

At work, that plays out in a familiar way. A product manager with hundreds of connections on a professional network site might talk to eight colleagues every week, have lunch with a handful of people from other teams each month, and see a former boss at one conference a year. The first group stays close almost automatically, the second holds steady only while the lunches continue, and the former boss drifts outward between conferences.

The outer circles still matter. Acquaintances you see rarely are often the ones who hear news your close circle does not, as the research on why acquaintances bring job leads your friends don’t shows, so decide which ties get regular time; there is no need to shrink your network to 150. A simple personal CRM, a list of people with the date you last spoke, makes that budget visible. A periodic network audit, mapping which parts of your working life have nobody in them, shows where the gaps are.

The table shows how much weight each study can bear.

The bottom line

Dunbar’s number is a rough estimate: the primate data behind 150 are too uncertain to fix any single figure, and its author still defends it with measured networks that average near that size. The part you can use is the time budget. Decide which few people get regular contact, expect friendships to fade when contact stops, and keep the wider network of acquaintances light.