Why Your Friend Group Shrinks After 25

In a 2013 study of 3.2 million European mobile phone users, the researchers Esteban Moro, Giovanna Miritello, and colleagues plotted call frequency against age and found a curve that surprised almost no one who'd been paying attention to their own life. The number of distinct people a person actively contacted by phone climbed steeply through the late teens, peaked in a sharp spike at roughly 25 years old, and then declined, more or less monotonically, for the next forty years. The peak network for a 25-year-old contained close to twice as many active contacts as the same person's network at 45. The pattern held for men and women, across countries, and the decline did not stop in middle age. By the late sixties, the typical network had contracted to a small, stable inner core.

The ceiling those networks bump into is named after Robin Dunbar, the British anthropologist who proposed it in a 1992 paper in the Journal of Human Evolution. Dunbar plotted average group size against neocortex ratio in 38 non-human primate genera, fit a regression line, and used the human neocortex to extrapolate. The prediction was a mean human group size of 148 — quickly rounded to 150 — with a 95-percent confidence interval running from 100 all the way to 230. Dunbar then went looking in the ethnographic literature and found that 150 kept reappearing: the splitting point of Hutterite agricultural colonies, the standard size of a Roman maniple and a modern infantry company, the size of a Neolithic farming village, the upper bound on the number of academics in any given research sub-specialty.

The number became corporate folklore through W. L. Gore and Associates, the Delaware-based company best known for Gore-Tex. The founder, Bill Gore, observed that once a single building exceeded about 150 employees, the social fabric tore in ways that hurt the work — people stopped recognising each other in the hall, factions formed, decision-making slowed. Gore's response was to cap each plant at 150 parking spaces. When the spaces filled, the company built a new building, often a few hundred meters from the last. Sweden's national tax agency reorganised on the same principle in 2007, capping office sizes at 150 employees and citing Dunbar by name in the planning documents. Malcolm Gladwell put the Gore example into general circulation in The Tipping Point in 2000.

Dunbar's later work added structure underneath the 150. He proposed a set of nested layers, each roughly a multiple of three larger than the one inside it: an inner core of about 5 — partners, parents, the people you'd call at three in the morning — wrapped in a sympathy group of 15, a friendship band of 50, and the active-acquaintance shell of 150. In a Financial Times piece in 2018, Dunbar wrote that humans devote about 40 percent of their available social time to that inner 5, and another 20 percent to the 10 people just outside it; two-thirds of all social effort goes to just 15 individuals. Beyond the 150, he added later, sit further layers — about 500 acquaintances and 1,500 recognisable faces — but those are passive recognition layers, not active relationships.

The number is contested. The American anthropologists H. Russell Bernard and Peter Killworth ran a series of field studies through the 1980s and 1990s using direct network-elicitation methods and arrived at a median of 231 active social ties — appreciably larger than Dunbar's figure, with a long upper tail. A 2021 paper in Biology Letters by Patrik Lindenfors and colleagues at Stockholm University re-ran Dunbar's regression with newer comparative phylogenetic methods and Bayesian statistics and got central estimates between 16 and 109, with confidence intervals so wide — 4 to 520 in one model — that the authors argued no single cognitive limit can be derived from the available primate data. Dunbar responded that the calculation should be restricted to apes, but that leaves a sample of only four group-living species.

What the call-record and questionnaire studies do agree on is the shape of the decay. The active network grows fast through adolescence as new social environments — high school, college, first jobs — pump in new ties, peaks shortly after the heaviest period of social mixing ends, then bleeds out in two distinct ways. The outer layers — the 150 and 50 shells — leak first, because they are maintained by frequency-dependent contact: a co-worker stops being a co-worker after a job change, a teammate disappears after the league ends, a downstairs neighbour fades after the move. The inner 15 are more durable but carry their own attrition: marriages, geography, illness, the deaths of parents. The total time budget for socialising stays roughly constant — most adults invest two to three hours of attention per day in other people — but the math forces a choice. Spending more on the inner 5 leaves less for the 50, and the 50 evaporates in months. The shrinkage isn't a personal failing or a generational symptom. It's what happens when a constant time budget meets a network whose inputs slow down after 25 and whose maintenance costs scale with frequency, not affection.