Summary: Thirty years ago, Robin Dunbar proposed that humans naturally maintain a core social group of about 150 people, including roughly five very close friendships. Despite repeated challenges, Dunbar’s Number remains supported by a growing body of evidence from anthropology, social data and neuroscience. Robin Dunbar explains why this pattern persists and how misunderstandings about statistics and social behavior have fueled criticism.
Source: The Conversation
Three decades ago, while examining a graph that related primate group sizes to brain size, I noticed a clear pattern: larger brains were associated with larger social groups. I wanted to know what group size that relationship would predict for humans.
My calculations pointed to about 150. At first I thought this might be too small, so I searched historical and ethnographic records for natural human group sizes. Hunter-gatherer societies consistently form nested social systems—families inside bands, bands inside communities, and communities inside tribes. The community layer, where people maintain meaningful social bonds, clustered around 150 individuals.
This observation led to the formulation of the social brain hypothesis—linking neocortex size to group size across primates—and the concept widely known as Dunbar’s Number, the approximate size of a meaningful human social circle.
Challenging Dunbar’s number
Because the idea attracted broad attention—from designers and architects to businesses and software developers—researchers have repeatedly questioned it. Scientific challenges are valuable when they clarify assumptions and reveal new insights, but many critiques of Dunbar’s Number have muddied the debate rather than advanced it.
Common objections argue that culture determines human social behavior, so biological constraints do not apply, or that digital platforms let people sustain far larger networks. These critiques often overlook a key point: Dunbar’s Number refers to the number of quality, emotionally meaningful relationships a person can sustain, not the much larger set of casual contacts or acquaintances that accumulate through social media.
A recent paper from Stockholm University claimed to refute Dunbar’s Number by suggesting the social brain equation underestimates human group size. However, that study relied on inappropriate statistical methods and did not take into account the broader empirical and theoretical evidence accumulated over the past three decades.
Developing Dunbar’s number
Over the last ten years, evidence for a layered social structure centered on roughly 150 has grown substantially. This pattern appears across diverse datasets: telephone call networks, online groups, Christmas card lists, military units, online gaming communities, church congregations, medieval village records and even Bronze Age settlements. The pattern is strikingly consistent.
Personal social networks and natural communities tend to display nested layers, each about three times larger than the one inside it. These same ratios appear in the multi-level societies of other large-brained mammals—monkeys, apes, dolphins and elephants—though humans show additional layers beyond those seen in other species.
Evidence from neuroscience
Neuroscience has helped fill in how the social brain supports these limits. Multiple neuroimaging studies have found that the size of an individual’s meaningful social network correlates with the size and connectivity of the brain’s default mode network, a system implicated in social cognition. In parallel, the neurochemical mechanisms that support social bonding—such as endorphin-mediated effects of touch and grooming—are shared between humans and other primates, which helps explain why physical contact and close social rituals remain important for maintaining relationships.
How not to do statistics
A principal flaw in the Stockholm study is a basic statistical error. The authors applied ordinary least squares regression (LSR) to predict human group size from brain size. LSR works well when the independent variable is measured precisely, but it underestimates slopes when both variables contain measurement error or when the relationship spans different grades of social organization.
The social brain relationship is better described by different grades across primate taxa. Using reduced major axis regression (RMA), which is more appropriate when both variables have error and when making predictive inferences, gives a very different prediction. For the same data, LSR yields a predicted human group size of about 71, while RMA and grade-appropriate models produce predictions near 150—consistent with the original estimate.
Beyond statistical technique, some critiques stem from incomplete familiarity with contemporary research on primate and human social systems. A careful reading of the accumulated literature clarifies both the limits and the explanatory power of the social brain framework.
About this social neuroscience research news
Author: Robin Dunbar
Source: The Conversation
Contact: Robin Dunbar – The Conversation
Image: The image is in the public domain