Your independent source for Harvard news since 1898

Your independent source for Harvard news since 1898


Networks, Neolithic to Now

May-June 2010

Networking” may be the defining buzzword for the early twenty-first century. But even before Facebook and LinkedIn existed, people had networks. In fact, a growing body of evidence shows surprising persistence in the size and quality of these networks—extending even into the online world: social-networking sites seem to reflect humans’ social tendencies, rather than change them.

A survey of all studies of group size in hunter-gatherer societies, assembled by Oxford evolutionary biologist Robin Dunbar, found that the average size of a clan or permanent village was 148 people. In a seminal journal article, Dunbar (a sometime collaborator with Nicholas Christakis, one of the Harvard scholars of networks profiled in “Networked") argued that throughout history, humans have kept coming back to this group size across the eons; this has come to be known as Dunbar’s number.

Dunbar cites findings that Neolithic villages in Mesopotamia typically had 150 to 200 residents and offers evidence that group size hovers near that same number for a wide range of contemporary cultures. These include rural communities in the Appalachian Mountains; subdisciplines in academia (when one grows larger than 150 members, a new subspecialty tends to branch off); and the Hutterites (who divide a farming community in two when it grows larger than 150 people, judging this the size at which controlling group members’ behavior through peer pressure alone becomes difficult; they prefer to split rather than create a police force). The unit size for most modern armies is also around 150—a number that has endured since the Roman era. “Given that the fighting power of a unit is a function of its size,” Dunbar writes, “we might expect there to be considerable selection pressure in favor of units that are as large as possible. That the smallest independent unit should turn out to have a maximum size of about 200 even in modern armies (where technology presumably facilitates the coordination of planning) suggests that this upper limit is set by the number of individuals who can work effectively together as a coordinated team.” (Christakis provides another modern-day example: his 2006 study of the freshman class at “a diverse private college in the northeast United States” found a very similar community size in the online world: the median number of Facebook friends among the students was 110.)

Dunbar concludes that this group size may have a biological basis: he analyzed the brain size and group size of various animals, yielding a formula that predicts a group size of 147.8 for humans. He suggests that for any organism, this number serves as “an upper limit on the size of groups that can be maintained by personal contact.” In fact, the structure of human social networks is remarkably similar across all societies that have been studied. But to date, there have been few such studies in traditional societies. Evidence that members of traditional groups form networks the same way would bolster the argument that humans’ friend-making patterns are hard-wired rather than culturally transmitted. Among those aiming to conduct such studies is anthropologist Coren Apicella, Ph.D. ’09, a postdoctoral fellow in the Christakis lab, who has conducted previous ethnographic research with the Hadza people of Tanzania and is now planning to map social networks among the Hadza or another traditional society. 


Social networks themselves may not be new, but a social-network perspective, combined with new tools, is yielding powerful new insights. University of Chicago political scientist John Padgett has developed a theory that upheaval in social networks in Renaissance Florence was responsible for the shift to modern capitalist and democratic societal structures. “The growth of trade with Asia caused some families to suddenly become wealthy, upsetting a feudal social network that was extremely hierarchical and disconnected between groups,” is how Christakis and University of California political scientist James H. Fowler ’92, Ph.D. ’03, summarize the theory in their book, Connected (Little, Brown, 2009). “New-money families started competing with old-money families for social control… At the center of the new social network was the Medici party, which spanned many of the previously disconnected groups. As a result, the Medicis were able to conquer once and for all the oligarchs who previously ruled Florence.” 


Christakis and Fowler are combining genetic information with new methods of network analysis to examine the role of genes in social behavior. Drawing on genetic and social-network data gathered in a study of 90,000 junior-high and high-school students around the United States, they compared data from the 1,110 sets of twins in the study with information on the non-twins in a 2009 paper that appeared in the Proceedings of the National Academy of Science. Comparing identical twins to fraternal twins, they found that genes accounted for 46 percent of the variation in how popular students were (i.e., how many other students named a given student as a friend), and for 47 percent of the variation in transitivity—the number of one’s friends who also know each other. A network with high transitivity is a tight-knit, insular group where everyone knows everyone; someone whose network has low transitivity has friends from many different groups, who in turn are unlikely to be friends with each other. In other words, if one twin was unpopular, the other twin was more likely than average to be unpopular, too; similarly, if one twin was popular, it was likely that the other twin was also well liked. (The researchers assume that the remainder of the variation in these characteristics is attributable to environment, which to some extent is shared between siblings whether they are twins or not.)

Christakis and Fowler are not surprised to find a strong genetic influence; they say it makes sense that people cobble together friendship networks that satisfy their partly genetically determined personality types. Explains Christakis: “Your genes can influence not just how many friends you have, but also your tendency to be in the middle or on the periphery of a social network—to pick popular or unpopular friends. And we think there is a genetic predilection to introduce your friends to each other: some people knit their networks together and some do not.”

 “On average,” they write, “a person with five friends who know one another has a different genetic makeup than a person with five friends who do not know one another.” They are continuing their research with studies of twin pairs on Facebook, comparing identical to fraternal twins on the basis of measures that include total number of friends and number of friends in common.

This is not the first research on genes and social behavior. In Connected, Christakis and Fowler note one finding of particular interest: a single gene variant distinguishes mating and parenting behavior in small mouselike mammals called voles. Male prairie voles, they write, “are paragons of monogamy, attaching to their first mate for life and taking care of their kids. Male meadow voles, in contrast, are much more promiscuous and less likely to care for their young. This stark difference in mating behavior suggests that evolution does not always yield behavior that we might consider moral in humans—sometimes it favors lust and the deadbeat dad. But more important, it shows what a big difference even a single gene can potentially make in the way animals connect to others.”

Christakis says that, to his knowledge, their paper is the first published anywhere on the genetics of human social networks. But the field is developing rapidly. Niels Rosenquist, another postdoctoral fellow in the Christakis lab, is guiding a set of studies on the genetic underpinnings of traits and behaviors that seem to be spreading through networks: obesity, alcoholism, smoking (see the main article and “Costs and Benefits of Connection” for more on this). Sorting out study subjects’ genetic propensity for such attributes may demonstrate that the network is actually more influential than previously thought—e.g., when your friend gains weight, you gain weight, even though he is genetically predisposed to obesity and you are not. Or it may find that part of what looked like a network effect is actually genetically based because people unknowingly choose friends with similar genes. For example, if a gene that favors obesity also confers a preference for a sedentary lifestyle, people with that gene variant may gravitate toward friends who are similarly slothful because they also have that allele.

Rosenquist, who has both a Ph.D. in economics and an M.D. degree, believes genetics will become increasingly important to the study of human social networks as genetic testing becomes cheaper, faster, and more accurate. Between genetic influences and network influences, he says, the entire concept of free will is being called into question, suggesting a new answer to the age-old question of what makes us do what we do: “The idea of the self is fundamentally changing….If we are essentially a function of our genes and our friends, what does that mean?” Rosenquist believes this area will develop rapidly, with heated debates that reflect the high stakes: here lies the potential to reach, for the first time, an understanding of human behavior that integrates the natural and the social sciences.

You Might Also Like:

You Might Also Like: