In 2013, an article appeared in the journal Social Forces with a striking title: “Breaking Up Is Hard to Do, Unless Everyone Else Is.” The authors, social scientists Rose McDermott, James Fowler, and Nicholas Christakis, had been studying how social phenomena spread. A professor at Brown, McDermott had noticed when going through a divorce that several couples whom she knew had also announced splits. Could this be more than coincidence? The authors set out to explore that question, using a data set of 5,000 people over a period of 32 years, from the famous Framingham Heart Study.
What they found seemed implausible: divorce is, well, contagious. It had been known for a long time that divorce “runs in families”—that is, divorced men and women are more likely to have parents who divorced, too. But these new results suggested that you can “catch” divorce, almost as you would the coronavirus. As with all viruses, when someone near us sneezes or coughs, we may not see the microbes that spread the nasty pathogen, but we know that they exist. Divorce contagion can’t be explained by a microorganism, yet McDermott and her coauthors stipulate that if your close friends divorce, your own chances of divorcing go up, and not by a little: by a full 75 percent. In fact, even if friends of your friends divorce, your chances of splitting rise, though to a lesser degree (33 percent). It doesn’t matter if those friends live in another state, or even another country. In a globalized, Facebook-connected world, the divorce “bug” easily moves across oceans.
We don’t need a psychologist to tell us that we often smile or laugh when others do. If you’ve gawked at the tiny humans in a neonatal nursery, you know how contagious crying can be. We use the phrase “going viral” to describe the latest pop or fashion hit for a reason. But divorce is a life-altering decision. People mull the pros and cons, the weight of which only they can determine; they toss and turn all night, worrying about their children, their finances, their homes, and their friendships. They consult counselors and lawyers and have heart-to-hearts with friends and relatives. And now some social scientists come along and say that, when you decide to divorce, you’re not an agonized individual weighing a moral choice but someone following a herd? “The idea insults the integrity of every person who finds themselves facing this incredibly difficult choice,” a Psychology Today columnist huffed after reading the Social Forces article.
It’s the strangeness of the idea of divorce contagion, though, that makes it such a vivid illustration of the power of human networks. As social animals, humans are linked to one another in myriad ways through families, neighborhoods, schools, workplaces, friends, and friends of friends. All these people, in turn, have their own connections, who, likewise, know and interact with their own networks, and so on across the globe. The science that studies these networks is evolving, and details remain hotly disputed. But spend time reading the research on networks, and no social phenomenon—from Bernie Sanders’s appeal to young voters to crime to school performance to poverty—looks quite the same.
Human networks are not random. We connect to others who’re like us in some way; social scientists call this homophily. A point of similarity might be something as fundamental to our identity as gender, religion, ethnicity, age, values, or beliefs. Or it could be, and often is, socioeconomic status. Or geographic—say, a neighborhood, city, or region where you grew up. Or an enthusiasm—football, a popular video game, or a breed of dog. When we meet someone at a party, we quickly try to find out if our networks connect in some way. Where are you from? What do you do? How do you know the host? Homophily is a human universal and an essential piece of our mental equipment for sustaining cooperation.
Though anyone can compile an inventory of friends and acquaintances, a network is not a mere sum of its individuals. It’s an array of interconnections, the characteristics of which themselves vary. Some ties are weak. Think of your friend’s law school buddy, whom you met a few times years ago but haven’t seen since. That tie probably has much less influence on you than, for instance, the best man or maid of honor at your wedding. Some ties are mutual, with the parties all reporting closeness; others are more one-sided. Networks can have a high level of “transitivity,” meaning that your friends are also friends with one another. Some are dense, when many potential contacts—in a neighborhood or school, say—are connected. Network scientists map these kinds of interconnections. To the untrained eye, their products resemble a two-year-old’s scribbles. But for the expert, the particular combination of strong and weak ties, transitivity, and the like reveals the architecture of a given group and its capacity for spreading a rumor of a friend’s engagement, a fad, the name of a new hip-hop recording—or divorce.
As it happens, we are getting a front-row view of social networks in action as the coronavirus advances across the globe. Humans are the perfect host for the virus to spread its DNA precisely because they are so stubbornly social. Now, with an abundance of cheap airlines, cruise ships, intercontinental businesses, and friendships, humans have extended their social reach just about everywhere on earth. The complex bundle of lines in the route maps you find in the glossy magazines of large airlines are one kind of network map, with enormous implications for social networks right now; the virus can easily hitch a ride from China to, say, Cincinnati. Once it arrives there, its host will offer it a chance to make the acquaintance of his or her family, friends, colleagues, and neighbors. With the help of census data on household makeup, workplaces, schools, and use of public transportation, network scientists can model “community spread” and recommend ways to reduce it. All those parents who have coped with closed local schools can fix their ire on a network scientist.
A virus spreads through social networks—that’s easy to grasp. Now, consider emotion, something far less obviously communicable. More than any other animal, humans have a large range of facial expressions, noises, and bodily postures—many subtle—that communicate feelings. I can see on a friend’s face an empty look in her eyes and the rigid set of her mouth typical of a depressed person. I spend time with her, hoping to cheer her up, but her depression may rub off on me. I then bring the blues home, putting my husband and children “at risk.” Contradictory as it sounds, social networks can actually spread loneliness or, more accurately, feelings of loneliness. Lonely people don’t necessarily have fewer connections. In fact, loneliness can appear in clusters, and un-lonely people will tend to get lonelier over time if they hang around lonely people. The mechanisms remain somewhat obscure. Maybe lonely people are less trusting and more anxious in ways that can influence others; maybe they stop returning phone calls or texts, weakening ties with friends and family.
“Humans are the perfect host for the virus to spread its DNA precisely because they are so stubbornly social.”
Interesting, yes, but none of this is particularly helpful for understanding how divorce spreads. For that, we need to recognize how networks evolve and reinforce rules about appropriate social behavior. Take obesity. In a 2007 paper, two of the authors of the contagious-divorce paper used the same Framingham data to collect information about participants’ weight. Sure enough: charting the networks revealed that obese people, like divorcées, appear in clusters.
Now that could merely be an example of homophily: an obese person, perhaps feeling self-conscious around those of average weight, might prefer to go to dinner and spend holidays with other obese people. Yet the theory wasn’t enough to explain the longitudinal data. It turns out that those with close ties to an obese individual—siblings, spouses, and good friends—were all more likely to become obese themselves, over the course of the study, than those with weaker connections. Could genetics be a factor? Unlikely: the effect was bigger on friends than on siblings. The less connected individuals are with obese people, the less likely they are to become obese. It was as if a network version existed of what medical researchers call a dose-response relationship. The higher the dose—in this case, exposure to obesity—the greater the response.
Why? The authors speculated that if you are of average weight but have a close relationship with someone obese, it alters your perception of normal body size—it re-norms your perceptions. Other research seems to support this. Several studies have shown that kids who grow up with obese parents and/or go to school with lots of obese children come to see larger body mass as not out of the ordinary. Another study, this one using data from Europe, confirmed that an individual’s view of proper weight depends on what his social network thinks. Highly educated people, who tend to assemble into homophilic networks, were more likely to think of themselves as overweight than were less educated people, even with the same BMI. The highly educated hold themselves to a thinner standard. It’s not obesity per se that’s contagious, in other words; it’s the idea of what constitutes obesity, and that varies by network. It seems likely that divorce contagion works similarly, with the acceptability of divorce depending on social connections.
Voting offers another example. Voting is a norm within many social networks; some will lie about having voted, just to avoid disapproval. Networks can also create voting cascades. Studies of “get out the vote” campaigns suggest that they have a modest impact; the person who answers the door is about 10 percent more likely to vote. That’s not the whole story, however. The new voter makes it more likely that her family and friends will also vote. “Instead of each of us having only one vote,” write Christakis and Fowler, “we effectively have several and are therefore much more likely to have an influence on the outcome of an election.” The authors calculate that one vote can “multiply” into three.
What’s both mysterious and disquieting about social-network theory is how it tampers with what we think of as our personal lives. Our thoughts and beliefs, though they feel unique to us, have a collective life. We’re used to thinking that what we wear, how we talk, whether we vote, what we eat, whether we separate from our spouse—all that is our own damn business. Our embeddedness in social networks suggests otherwise. Like it or not, we’re all in one another’s business.
Network science can be exceedingly technical, with its complex graphs and alien vocabulary of transitivity and centrality. Few people outside the universities, public-health organizations, and marketing firms have much reason to encounter its lessons. Still, some of its basic insights should be evident to one group above all: parents.
When parents complain about “peer pressure,” or when they refer to a child’s classmate as a “bad influence,” they’re really talking about social networks. Children sort into friendship groups as early as preschool, but American parents especially start worrying when their kids reach their teens. During those years, we’re often told, children become more independent as they try to establish an adult identity. But the word “independent” is misleading here. As they move into middle and high school, children move around more independently and might keep their bedroom door shut for suspicious lengths of time—but they develop their identity largely through their friends, not from individual self-reflection.
Just like adults, kids assemble themselves into groups of people like themselves. But their groups tend to differ from those of adults in crucial respects. Adolescents create densely knit networks. Their friends will disproportionately be friends with their friends’ friends; their networks are high in transitivity. This deepens group cohesion, which, in turn, helps to reinforce norms as well as to introduce irritating fads. Partly, this results from the closed ecosystem of a high school. Adolescents back in the day, working on a farm or apprenticing for a tailor, didn’t interact constantly with intertwined, same-aged peers the way modern adolescents do. And while they may have endured hours of tedious work and bruising orders from coldhearted bosses and feared stigma within their communities, they would have been relatively free of chronic peer pressure. Peer pressure is a major negative by-product of the modern high school. If today’s parents should be worried about something, it’s not adolescent rebellion; it’s adolescent conformity.
Network research reinforces our intuitions about peer pressure. Yes, kids attach themselves to peers similar to themselves in some ways; at the same time, those friends influence them. Middle schoolers are more likely to smoke pot if their friends do. They’re more likely to start having sex if their friends are. Worried about your son or daughter binge-drinking? Find out if your teen’s friends are getting wasted. And watch what the popular kids—the celebrities of high school—are doing. Just as knockoffs of a dress worn by Kate Middleton sell out overnight, so word of a popular high school student chugging a six-pack will make it seem like the cool thing to do. A new report out of Sweden tells us that the country saw a 1,500 percent rise in gender dysphoria in adolescent girls between 2008 and 2018. The phenomenon hasn’t been studied yet, but it seems like a textbook case of social contagion, doubtless helped along by a few popular teens.
High schools are as hierarchical as the African savanna. Everyone knows who the alpha-popular kids are; in social-network terms, they’re the teens with the most connections to the kids with lots of connections. If they undertake a new behavior—smoking, say, or coming out as trans—it makes that behavior appear more widespread than it is, a dynamic that social media only intensifies. In his book The Human Network, Matthew Jackson observes that, on Facebook and Instagram, the popular teens have the most followers. If they get a tattoo, many in their network will see it; some will then get one for themselves. Popular adolescents also attend a lot of parties where smoking, drinking, and drug use might be more expected. Few post pictures of themselves reading a book. “It is thus natural,” writes Jackson, “for a teen to overestimate the amount of drugs and alcohol consumed by his or her peers and to underestimate the amount of time spent studying by those same peers.”
A far more disturbing network effect emerged in a rash of adolescent suicides in a wealthy suburban community beginning in 2005. Sociologists had long known that suicide can happen in clusters, and they felt that it wasn’t simply because of “peer selection”—that is, of suicidal individuals seeking out one another. Instead, researchers speculated, social networks may have accelerated an outbreak of suicide. University of British Columbia sociologist Seth Abruytin and his coauthors undertook a qualitative study of the affected community—dubbed “Poplar Grove”—to determine what psychological mechanism might trigger such extreme behavior.
What they found was a particularly toxic variation of the way that networks spread norms. As the community struggled to cope with the tragedies, a theory arose about their cause. Academic pressure was to blame, the theory held—and it soon became widely accepted. But regardless of how the academic-pressure explanation might apply in particular cases, it had the unintended effect of creating a “normalized” script for the more extended network. “The new suicide script,” the authors wrote in a 2019 article in Society and Mental Health, “reinterpreted broadly shared adolescent experiences (exposure to pressure) as a cause of suicide, facilitating youth’s ability to imagine suicide as something someone like them could do to escape.” That the dead included some popular teens, who’d exhibited no clear signs of depression, suggested the intensity of the influence.
The dynamics of contagion don’t have to be destructive, though. They can also lead a teen to study for exams, to get higher grades, and to go to college. Kids may be influenced to start smoking because their friends do, true; but they also might decide to stop because their friends have stopped. In some environments, the popular kids are the biggest rebels and risk-takers; in others, they are the aspiring Ivy Leaguers. The right social network can create a kind of herd immunity to outlier behavior. A network can shame the meth-taker or macho bully either to stop the undesirable action or to look for a different network, where those behaviors are more accepted.
Ensuring that their children grow up in an environment with more Yale wannabes than meth addicts occupies a big chunk of many parents’ energies. Making sure their sons or daughters are hanging around the right peer group is a big reason that they’re willing to risk bankruptcy to buy a home in Great Neck or to write sweat-inducing checks for Sidwell Friends. “Proximity has a huge influence on our friendships,” Jackson notes. Parents know it.
Network science, then, raises a troubling question. Affluence cannot fully protect kids from even the most toxic network effects: the Poplar Grove suicide epidemic proves that. Still, that’s an extreme case. Middle-class parents generally have the means to see to it that their children are surrounded by peers who behave reasonably well in class, do their homework, smoke weed only occasionally (if at all), and have never seen the backseat of a police car or had a positive pregnancy test. But what about children whose parents can’t afford to curate their kids’ networks in such a fashion, or otherwise lack the wherewithal to do so? What about parents who don’t have much choice but to live in high-poverty neighborhoods with high-poverty schools, both crowded with troubled peers?
One way to think about the origins of delinquent adolescent behavior is to take a close look at the offender’s home. “Our mothers all are junkies / Our fathers all are drunks / Golly Moses, natcherly we’re punks,” as the Jets wisecrack to Officer Krupke in West Side Story. The statistics support this diagnosis, but as an explanation for crime, substance abuse, school failure, and unemployment, individual psychology and family background have their limits.
Think of how often we hear locals protest, after a neighborhood teen gets accused of a crime: “But he was from such a good family!” In some instances, it’s true. In schools where enough peers take drugs, commit crimes, and drop out of school and the labor market, some percentage of children raised in loving, orderly homes will also become “punks.” Poor neighborhoods where domestic and street violence are common are unlikely to be friendly places for children worried about the next chemistry exam. In those environments, aggressive, rule-defying teens, confident enough to swagger along the mean streets, are the popular ones, or, in network terms, the ones with high connectivity. That’s why we see so many families clamoring for a spot at charter schools like Success Academy. Parents aren’t just looking for better teachers and curricula for their children; they’re also trying to get them away from toxic networks.
In an influential paper, Stanford sociologist Mark Granovetter argued that, though ties with family members and close friends exercise a powerful influence on our norms, weak ties—our acquaintances who generally don’t know one another—remain crucial for getting ahead in a complex society. The poor tend to have networks “rich in strong ties but poor in weak ties”: think of blood-oath gangs or residents of small, isolated villages. These ties foster loyalty and conformity. But it’s the weak ties with people whom they don’t know well—say, someone their sister once worked for, or a Little League coach from years back, now living in a different town—that communicate information about schools and jobs, as well as alternatives to maladaptive local norms. Especially in a large, mobile society like ours, individuals thrive by being embedded in a variety of social networks. They establish different networks every time they move—and for each job held, school attended, and organization joined. The poor, researchers find, have a more limited range of ties, and often meet no one outside their immediate network. While dense networks hold members in a reassuringly tight embrace, it’s weak ties that provide a bridge to a social world beyond.
In sum, social networks can reinforce the disadvantages of the poor as well as the advantages of the well-to-do. They shore up inequality not just because, say, a friend’s cousin is on the board of a top college or a neighbor works at a publishing firm offering summer internships but also because of the way they spread information and buttress norms. It’s not a “culture of poverty that is the problem; it is the social network of poverty,” sociologist Charles Kadushin observes.
This sounds like a pretty strong argument for greater integration. Take children from troubled neighborhoods and their failing schools and put them into communities with affluent kids and high-performing schools, goes the thinking, and they’ll model themselves after their high-achieving peers.
There’s some evidence for this claim. A 2019 working paper from the Philadelphia Federal Reserve, for instance, found that gentrification increases the probability that children of less educated homeowners will attend, and graduate from, college. The authors attribute the gains to the fact that children see wealthier and more educated neighbors around them and enjoy more advantageous networks. Perhaps the strongest case comes from Moving to Opportunity, a 1990s federal experiment that randomly assigned 4,600 families in five American cities living in high-poverty housing projects to three groups: one that received a housing voucher, with a proviso that it be used in a low-poverty neighborhood; and two control groups. The early results were a dud, showing no improvement in children’s academic performance or behavior (in fact, boys seemed worse off) and no change in adult employment or earnings. In 2016, however, a team led by Harvard economist Raj Chetty looked at more recent data and discovered better news. They confirmed the earlier researcher’s conclusions that moving adults into thriving neighborhoods had no effect on their employment or earnings and found few benefits for children who moved as teens. The big story was younger children. Kids who moved before age 13 saw a 31 percent rise of adult annual income, and total lifetime earnings of about $302,000, as well as slightly higher rates of marriage and college attendance. The authors don’t find a critical age for moving, but “every extra year of childhood spent in a low-poverty environment appears to be beneficial.”
As policy interventions go, these are dramatic results. Unfortunately, other research should temper enthusiasm. According to a number of studies, once a school moves beyond a tipping point of background diversity—20 percent to 30 percent—students start to self-sort into networks according to race, ethnicity, and socioeconomic status; and they gravitate toward classmates with a similar level of academic achievement. Educators can get rid of tracking to try to ensure that lower-achieving children rub shoulders with higher-achieving classmates, but the kids still track themselves. Whatever educators do, the high-achievers like to hang out together; the same goes for the low-achievers.
The birds-of-a-feather instinct runs so deep that children start sorting themselves as early as preschool. For a 2016 study, researchers watched a group of Head Start children interacting with classmates over the course of a year. The kids—four-year-olds, mind you—gravitated toward others with similar “preschool competency,” meaning that children who actively participated in class and listened to teachers grouped together. Elementary school children are drawn to same-race friends. The heterophilic friendships they do have are more likely to break off before fifth grade.
People engage in similar self-sorting at the neighborhood level. Urban planners have long preferred mixed-income, mixed-race neighborhoods, partly under the assumption that they would encourage mingling, tolerance, and economic mobility. Sociologists don’t find much of that. One study of Boston’s South End, which subjects described as the “most diverse neighborhood in New England,” with “million-dollar homes alongside subsidized housing,” found that residents with different backgrounds joined different civic groups and made use of different public spaces. Other studies find that residents of the same areas, depending on their backgrounds, use separate stores and beauty salons; two restaurants can be right next to each other, and they’ll still have completely different customer bases. And this is not just an American phenomenon.
Homophily will never disappear, and we shouldn’t hope that it will. It is a crucial social glue, but it also limits the information we’re likely to receive and our understanding of outsiders. It is congenial to friendship but hostile to diversity, as the chances of misunderstanding, clashing norms, and mistrust inevitably rise.
Social networks can’t lead us to any ultimate Grand Theory of Human Social Behavior. They can’t explain why some people are more susceptible to influence, both positive and negative, than others. All of us encounter friendships and marriages that defy any obvious form of homophily; opposites do seem to attract at times. Homophily also fails to account for how charity and kindness to strangers from dramatically different cultures and religions energize major philanthropies and modest church tithing alike. There are depths to the human imagination that the scientist can’t plumb.
Still, policymakers and educators might be able to mine some insight from social-network theory. That’s already been the case for a while in the field of public health; the ubiquitous charts and graphs of the coronavirus’s spread and predictions for the future are an unwelcome, though ubiquitous, reminder of the reach of that discipline. Educators are now experimenting with targeting more popular students with “pro-social” messages about smoking, early sexual activity, and the like.
The problem is that trying to manipulate social networks, as in the last example, will inevitably have consequences that we can’t foresee. Could recruiting popular students to spread desired messages further surrender adult authority to the already-too-powerful peer group or prop up noxious social hierarchies? Scientists may have something to teach us about the way norms spread; they can’t tell us what norms we should value or describe how institutions like schools should best communicate those norms. Yes, we are creatures of contagion, pushed here and there by mysterious, unknowable forces. Especially at this moment, understanding that truth is essential. It should also be humbling.
Photo: Petrut Calinescu/Alamy Stock Photo
