Contagious Behavior: How We Influence Each Other

The old saying, “Tell me who your friends are, and I’ll tell you who you are,” may hold more truth than we usually think. Not only do our closest friends shape who we are, but so do our friends’ friends: they can help us quit smoking or contribute to us gaining weight, make us happier or lonelier. To be fair, we also influence people we may not even know directly. This article, based on Clive Thompson’s piece for The New York Times, explores research and criticism around the theory of social connections and “contagious” behavior.

The Framingham Story: How Our Social Circles Shape Us

Seventy-four-year-old Eileen Belloli makes an effort to maintain her friendships. Born in Framingham, Massachusetts, she met her future husband, Joseph, there. Both have never left Framingham, just like many of Eileen’s friends from elementary school, so even after 60 years, they still gather every six weeks.

Last month, I visited the Belloli family and asked Eileen about her friends. She immediately pulled out a folder with photos from school years and class reunions. Every five years, she helps organize a reunion, and they always manage to gather about 30 people. Flipping through the photos, I could see that the Bellolis and their friends maintained good health for many years. As they aged, they mostly stayed slim, even though many other Framingham residents died from obesity. Eileen is especially proud of staying active. Her only vice was smoking: after her workday as a biology teacher, she would go to a nearby café, have two cups of coffee, and smoke two cigarettes. At the time, her smoking habit didn’t seem like a problem—most of her friends smoked too. But in the late 1980s, some began to quit, and soon Eileen felt awkward holding a cigarette. She quit as well, and within a few years, no one in her circle smoked anymore.

There was only one person in the reunion photos whose health noticeably declined over the years. When he was younger, he looked as healthy as everyone else, but year by year, he grew larger. He didn’t stay friends with his classmates; his only connection to them was these reunions, which he attended until last year. Later, it turned out he had passed away.

The Framingham Heart Study and Social Contagion

This man’s story is especially relevant because Eileen and Joseph participate in a scientific study that may help explain his fate. The Framingham Heart Study is the most ambitious national project on the causes of heart disease, started in 1948 and covering three generations of families in the town. Every four years, doctors examine every aspect of participants’ health—heart rate, weight, cholesterol, and more. For decades, the study was a goldmine of information about heart disease risk factors.

But two years ago, sociologists Nicholas Christakis and James Fowler used the data collected on Joseph, Eileen, and thousands of their neighbors to make a groundbreaking discovery. According to Christakis and Fowler, analyzing the Framingham data gave them the first solid foundation for a powerful theory of social epidemiology: good behaviors—like quitting smoking, staying positive, or keeping slim—spread from friend to friend almost like contagious viruses. The same was true for bad behaviors: groups of friends seemed to “infect” each other with obesity, unhappiness, and smoking. It appears that good health is not just a matter of genes and diet, but also partly a result of being close to other healthy people.

The Roots of Social Contagion Theory

For decades, sociologists and philosophers suspected that behavior could be “contagious.” In the 1930s, Austrian sociologist Jacob Moreno drew sociograms—small maps of who knows whom—and found that social connections vary widely. Some people were “stars,” chosen by many as friends, while others were “isolates,” with few or no friends. In the 1940s and 1950s, some sociologists analyzed how the shape of a social network could influence behavior, while others studied how information, gossip, and opinions spread within these networks. One pioneer, Paul Lazarsfeld of Columbia University, analyzed how commercial products become popular. He argued that popularity grows in two steps: well-connected people first absorb media advertising, then tell their many friends about the product.

Today, we often talk about social change as an epidemic (like the “obesity epidemic”) and about “super-connectors” who have such strong ties that they almost single-handedly drive trends. However, in none of these studies did researchers observe the process of “infection” directly. They reconstructed it after the fact, often relying on interviews, which are prone to memory errors. Plus, these studies focused on small groups (a few hundred people at most), so they might not reflect how “contagious” behavior spreads—if it does—across the general public. Are “super-connectors” really that important? How many times does someone need to encounter a trend before “catching” it? While scientists knew people could influence close colleagues, could that influence spread further? Despite belief in social contagion, no one really knew how it worked.

Christakis and Fowler: Mapping Social Influence

Nicholas Christakis developed a new perspective on this question in 2000 after visiting terminally ill patients in Chicago. Christakis—a physician and sociologist from Harvard—had made a name for himself studying the “widowhood effect,” the tendency for spouses to die soon after their partners. One of his patients was an elderly woman with dementia, cared for by her daughter. The daughter was exhausted, her husband became ill from stress, and then a friend of the husband called Christakis for help, saying he too felt depressed by the situation. The illness of one woman had spread “three degrees of separation”: to her daughter, her son-in-law, and his friend. This made Christakis wonder how to study such phenomena further.

In 2002, a mutual friend introduced him to James Fowler, then a Harvard political science grad student. Fowler was researching whether voting decisions could spread virally from person to person. Christakis and Fowler agreed that social contagion was an important field and decided the only way to answer the many open questions was to find or collect a huge dataset covering thousands of people. They considered running their own study but instead searched for existing data. They weren’t optimistic: while there are large health surveys, medical researchers rarely ask who knows whom among their patients.

But the Framingham study looked promising: it had run for over 50 years and tracked more than 15,000 people across three generations. Theoretically, it could provide the needed picture—but how to track social ties? Christakis got lucky. During a visit to Framingham, he asked a study coordinator how they kept in touch with so many people for so long. She pulled out a green sheet—the form used to collect information from each participant at every checkup. It asked about spouses, children, parents, siblings, where they lived, who their doctor was, where they worked and lived, and who their close friend was. Christakis and Fowler could use these thousands of green forms to manually reconstruct Framingham’s social network over decades.

Over the next few years, the researchers led a team that painstakingly reviewed the records. When finished, they had a map of how 5,124 subjects were connected: a network of 53,228 ties among friends, families, and colleagues. They then analyzed the data, starting with patterns of weight gain, and created an animated diagram of the entire social network, where each resident was a dot that grew or shrank as the person gained or lost weight over 32 years. The animation showed that obesity spread in clusters. People didn’t just gain weight randomly.

The social effect was powerful. When one Framingham resident became obese, the likelihood of their friends becoming obese increased by up to 57%. Even more surprising, the effect didn’t stop there: a resident was about 20% more likely to become obese if a friend of a friend did—even if their own close friend stayed the same weight.

“You may not know him personally, but your friend’s husband’s coworker can make you gain weight. And your sister’s friend’s boyfriend can make you slimmer,” Christakis and Fowler would later write in their book Connected.

Obesity was just the beginning. Over the next year, the sociologist and political scientist continued analyzing Framingham data, finding more and more examples of contagious behavior. Drinking, happiness, and even loneliness spread through society in similar ways. In each case, individual influence extended up to three degrees before fading. The researchers called this the “three degrees of influence” rule: we are connected not only to those around us, but to everyone else in this web, which stretches much farther than we think.

How Does Contagious Behavior Actually Spread?

But how could obesity or happiness spread across so many links? Some contagious behaviors, like smoking, are easy to explain: if many people around you smoke, you’ll feel peer pressure, and if no one smokes, you’re more likely to quit. But peer pressure doesn’t explain happiness or obesity: we don’t often urge people to eat more or be happier.

To explain the phenomenon, Christakis and Fowler hypothesized that such behaviors partly spread through subconscious social cues we get from those around us, which serve as hints about what is considered normal in society. Experiments show that if someone sits next to a person who eats more, they will also eat more, unconsciously adjusting their idea of a normal meal. Christakis and Fowler suspect that as friends around us gain weight, we gradually change our idea of what “obesity” looks like and silently allow ourselves to gain weight. In the case of happiness, they argue that contagion may be even more deeply subconscious: the spread of good or bad feelings may be partly caused by “mirror neurons” in our brains, which automatically mimic what we see on others’ faces.

This subconscious emotional mirroring may explain one of the study’s most curious findings: if you want to be happy, the most important thing is to have lots of friends. Historically, we’ve thought that having a small group of close, long-term friends is key to happiness. But Christakis and Fowler found that the happiest people in Framingham were those with the most connections, even if the relationships weren’t deep. The reason is likely that happiness doesn’t just come from deep conversations—it also comes from encountering many small moments of contagious happiness from others every day.

Of course, the risk of having many close ties is that you may encounter more people in a bad mood. However, being more social always pays off for one surprising reason: happiness is more contagious than unhappiness. According to the researchers’ statistical analysis, each additional happy friend boosts your mood by 9%, while each unhappy friend drags you down by only 7%.

The Framingham study also suggests that different contagious behaviors spread differently. For example, coworkers, unlike close friends, don’t transmit happiness to each other, but they do transmit attitudes toward smoking. Obesity had its own pattern: spouses didn’t influence each other as much as friends did. If a male Framingham subject had a male friend who gained weight, his risk doubled, but if his wife gained weight, the risk increased by only 37%. This may be because, when it comes to body image, we compare ourselves mainly to people of our own gender (and in the Framingham study, all spouses were opposite-sex). Similarly, friends of different genders didn’t transmit obesity at all: if a man became obese, his female friends weren’t affected, and vice versa. Same-gender siblings (two brothers or two sisters) influenced each other’s weight more than opposite-gender siblings (brother and sister).

When it came to drinking, Christakis and Fowler found a different gender effect: women in Framingham were much more influential than men. A woman who started drinking heavily increased the risk of alcohol use among those around her, while drinking men had less influence. Fowler believes women have more impact because they usually drink less, so when a woman starts drinking heavily, it’s a strong signal to others.

Criticism and Alternative Explanations

The researchers’ work sparked a range of reactions. Many public health experts were thrilled. After years of observing patients, they suspected that behavior patterns spread in society, but now they had data to back it up. However, many network scientists were more cautious. Unlike medical experts, these researchers specialize in studying networks themselves—from power grids to teens on Facebook—and are well aware of the difficulty in establishing cause and effect in such complex systems. They note that the Framingham study found intriguing correlations in people’s behavior, but that doesn’t prove social contagion causes the spread of a phenomenon.

There are at least two other possible explanations. One is “homophily”—the tendency of people to gravitate toward those like themselves. People who gain weight may prefer to spend time with others who are also gaining weight, just as happy people may seek out other happy people. The second possible explanation is that a shared environment—not social infection—may cause Framingham residents to share behaviors within groups. If a McDonald’s opens in a Framingham neighborhood, it could cause a group of nearby residents to gain weight or become happier (or sadder, depending on their view of McDonald’s).

One of the most vocal critics of Christakis and Fowler is Jason Fletcher, associate professor of public health at Yale. He and economist Ethan Cohen-Cole published two papers arguing that Christakis and Fowler didn’t account for all possible homophily and environmental effects. Fletcher wanted to repeat their analysis but didn’t have access to their data. Instead, he and his colleague tested Christakis and Fowler’s statistical methods on another dataset—the Add Health study, a federal project tracking the health of 90,118 students in 144 middle and high schools from 1994 to 2002. Students were asked to list up to 10 friends, allowing Fletcher to map out social networks in each school and test the math. When Fletcher analyzed the forms using similar statistical tools, he found that social contagion did exist, but the behaviors and conditions that were “contagious” included things like acne, height, and headaches. How can you become taller by hanging out with taller people? This, Fletcher concluded, cast doubt on whether Christakis and Fowler’s methods really ruled out homophily or environmental effects, and thus whether the Framingham results were reliable.

Fletcher said he believes in the reality of social contagion, but Christakis and Fowler’s evidence simply doesn’t impress him.

Other scientists pointed out another limitation: the Framingham network map is necessarily incomplete. When participants were checked every four years, they were asked to list all family members but only one close friend. This could mean the three-degree influence effects were an illusion.

When I raised these concerns with Christakis and Fowler, they agreed their friendship map was imperfect, but said they believed it had fewer holes than critics claimed. When they tallied the “green sheets,” they often found relationships between two people who hadn’t listed each other, reducing the number of false three-step links.

They also admitted it’s impossible to completely rule out homophily and environmental effects, but that doesn’t mean they agree with Fletcher. Christakis and Fowler point to two other findings to support their case for social contagion over environmental influence. First, in the Framingham study, obesity could spread from person to person even over long distances. When people moved to another state, their weight gain still affected friends back in Massachusetts. In such cases, the local environment couldn’t be responsible for both people gaining weight.

The second finding is even more intriguing: they found that behavior seemed to spread differently depending on the type of friendship. In the Framingham study, people were asked to name a close friend, but friendship wasn’t always mutual. While Steven might name Peter as a friend, Peter might not feel the same about Steven. Christakis and Fowler found that this “directionality” mattered: if Steven became obese, it didn’t affect Peter, because Peter didn’t consider Steven a close friend. But if Peter gained weight, Steven’s risk of obesity nearly doubled. If two men considered each other friends, the effect was huge: one gaining weight almost tripled the other’s risk. Christakis and Fowler found this directionality effect even among people who lived and worked very close to each other. This, they argue, means people can’t become more obese just because of the environment, since the environment should affect everyone equally, but it didn’t.

The directionality effect seems significant, supporting the argument for social infection.

Implications for Public Health and Personal Life

Essentially, Christakis and Fowler’s work offers a new perspective on public health. If they’re right, public health initiatives that focus only on helping those already affected are doomed to fail. To truly fight the spread of bad social behaviors, you must also focus on people so distant they don’t even realize they’re influencing each other.

It’s tempting to think, after reading Christakis and Fowler, that the best way to improve your life is to cut ties with people who have bad habits. And that’s possible—people do change friends, sometimes abruptly. But changing your social network may be harder than changing your behavior: research shows we don’t have as much control as we think over how we’re connected to others. For example, our position in the social network, or how many of our friends know each other, are relatively stable patterns in our lives.

Christakis and Fowler first noticed this effect when studying happiness. They found that people deeply embedded in circles of friendship were much happier than “isolates” with few connections. But if an “isolated” woman did become happy, she didn’t suddenly gain new connections or move into a more connected position. The reverse was also true: if a well-connected person became unhappy, they didn’t lose their connections or become “isolated.” In other words, your place in the network affects your happiness, but your happiness doesn’t affect your place in the network.

Are We Truly Independent Individuals?

The science of social networks ultimately offers a new perspective on an age-old question: to what extent are we independent individuals?

Viewing society as a social network, rather than just a collection of people, can lead to some thorny conclusions. In a column for The British Medical Journal, Christakis wrote that a strictly utilitarian view would suggest we should provide better medical care to well-connected people, since they’re more likely to pass those benefits on to others. “This conclusion,” Christakis wrote, “troubles me.”

Still, as the two researchers argue, there’s something inspiring in the idea that we’re so closely connected. “Even if we’re influenced by others, we can influence others,” Christakis told me. “And so the importance of doing things that benefit others increases. The network can work both ways, undermining our free will, but also, if you like, increasing the importance of having free will.”

As Fowler noted, if you want to make the world better through your good behavior, the math is on your side. Most of us are connected within three degrees to more than 1,000 people—these are all people we can theoretically help become healthier, more energetic, and happier simply by our own remarkable example.