Drake Bennett in Businessweek takes on evolutionary anthropology this week in a profile of Robin Dunbar ("The Dunbar Number, From the Guru of Social Networks"). If you don't know why Businessweek would be profiling a primatologist, then you probably don't know about Robin Dunbar's work with group size and communication.
Dunbar proposed that the size of social groups is limited by our evolutionary history. Our social interactions require us to track social relationships with many people. People aren't capable of following the details of thousands of city-dwellers, we are better suited to follow details of a single city block. In Dunbar's model, we should be well adapted to track the number of people that would have occupied social groups in the distant past.
Humans can keep track of only a limited number of people, and for most social primates the group size is even smaller. Dunbar's research on primate groups led him to believe that group size is correlated with brain size among species of social primates. Given this, we might expect that the exceptionally large size of the human brain would correspond to an exceptionally large group size. By drawing a regression among brain size and group size estimates for many primate species, Dunbar arrived at the prediction that human group size should be 150 people. That became known as the "Dunbar number".
Later, as the article recounts, technologists interested in social networks such as Facebook and Twitter became interested in the concept. What if the sizes of social networks on these alternative media platforms is similarly limited by our evolutionary heritage? Surely there is big money to be made for the company that can use this evolutionary knowledge to make big money?
But the naive understanding of a hard "number" limiting human social interactions doesn't fit the evolutionary evidence.
Others, anthropologists and brain scientists in particular, challenge the evolutionary story Dunbar tells, arguing that it discounts other factors that might have driven the development of the big human brain—the pressure to figure out more efficient ways to forage, or the need to surmount the defense mechanisms of the plants and animals our ancestors wanted to eat. “Ecological pressures like avoiding predators, finding food and shelter, choosing habitats—all these kinds of decisions. I think they played a role” in brain growth, says Reader, the biologist.
Researchers who’ve used different methods to measure the size of a person’s social circle have come up with numbers that don’t match Dunbar’s. One set of studies by the anthropologist Russell Bernard and the network scientist Peter Killworth found a mean social network size of 291. Another paper, published this month in the Journal of the American Statistical Association, came up with 611.
Like many, I used to be annoyed about the idea of the "Dunbar number". Human hunter-gatherers have a huge range in actual group sizes, as do most social primates. A girl might be born in a group of 20, see that group swell or merge with other groups to 100 or more, transfer to a different group along with a sister, and marry outside her social group entirely. Over a lifespan, she may face many different group sizes, and have "tight" relationships with other individuals born more than a century apart.
No single number could describe this complexity. People vary extensively in their social interactions and knowledge of other peoples' social lives. Some people track hundreds of others, some people are relative shut-ins and know only a few. Human environments are novel today in many ways relative to our evolutionary history, but the sheer scope of variation in sociality belies the notion that we come from a monolithic history. It makes about as much sense as saying that human mass is 65 kg and not recognizing the development (from 3 kg at birth) and extensive variation in mass among individuals.
My attitude has softened toward the Dunbar number, because it does capture something about the scale of social networks. Humans are not all 65 kilograms, but more than 99 percent of adult humans are within factor of two of that. Most of the time, when articles describe the concept of the Dunbar number, they describe it in ways that signify scale rather than an exact number. From this article, for example:
A paper published in 2011 found that on Twitter the average number of other people a user regularly interacts with falls between 100 and 200. And though the limit on how many Facebook friends one can have is a generous 5,000, the average user has 190—more than 150, but within what Dunbar sees as the margin of error.
If 190 is within the margin of error of 150, we're talking about a scale within a factor of at least 1.3. And that's on a mean -- not on the actual variation.
Cognitive limits do constrain our behavior. But species-wide cognitive limits have almost no ability to predict the sizes of particular groups. Hence orangutans, gorillas and chimpanzees have very different social group sizes despite similar brain sizes. And cognition isn't dedicated to a single social function; there are many:
Morin likes to point out that it’s misleading to talk about a single Dunbar Number. Dunbar actually describes a scale of numbers, delimiting ever-widening circles of connection. The innermost is a group of three to five, our very closest friends. Then there is a circle of 12 to 15, those whose death would be devastating to us. (This is also, Dunbar points out, the size of a jury.) Then comes 50, “the typical overnight camp size among traditional hunter-gatherers like the Australian Aboriginals or the San Bushmen of southern Africa,” Dunbar writes in his book How Many Friends Does One Person Need? Beyond 150 there are further rings: Fifteen hundred, for example, is the average tribe size in hunter-gatherer societies, the number of people who speak the same language or dialect. These numbers, which Dunbar has teased out of surveys and ethnographies, grow by a factor of roughly three. Why, he isn’t sure.
One way to look at it is scaling: Larger groupings are in some ways metagroups: composed themselves of smaller groups that can act semi-independently. These dynamics are limited in scope before the metagroups disintegrate under their own complexity.
But I think the null hypothesis is that we are imposing this scaling on the data. A factor-of-three difference is just wide enough for an analyst to separate biological distributions without much overlap. Try to sort human activities into different sizes of groups, and you'll get a factor-of-three separation automatically.
Are different "levels" of social relationship really adaptively different? If they are, we might be drawing upon different cognitive resources for different levels of relationship -- one kind of thinking for close friends, another kind for distant acquaintances. If instead the appearance of different levels is really just a consequence of mapping a continuous distribution into different sizes of groups, then the levels may not be adaptively different, and we may be using the same cognitive resources in ways scaled to our knowledge of the individuals involved. That adaptive question has to shape the way we conceive of limits on cognition, and the relationship between cognitive evolution and social dynamics. It will be more complicated than a single number.