Connected: The Amazing Power of Social Networks and How They Shape Our Lives

This diagnosis did not sit well with the community, and it upset many of those who had been ill, such as one twelfth grader who was quoted as saying, “They said we were crazy…. It just made me mad. When I’m sick, I don’t want someone to say I’m faking. They wouldn’t have taken me to the hospital, and my blood pressure wouldn’t have been sky-high, if I wasn’t sick.”

(#litres_trial_promo) Of course, the symptoms of those with MPI. whether laughing, dancing, fainting, or nausea, are quite real; they do not “fake” their experience in the deliberate, premeditated way that a malingerer does. The astonishing reality is that our own anxiety makes us sick, but so does the anxiety of others.

The CDC investigators also discussed why communities tended to use so many resources to try to find environmental causes for conditions that appeared to be psychogenic. The problem is that while public health professionals often suspect that an outbreak is psychogenic, they feel they have no choice but to conduct an unreasonably thorough investigation because of intense anxiety in the community. And, of course, it is very difficult, if not impossible, to definitively prove that a mysterious toxic exposure has not simply escaped detection. The CDC investigators noted the possibility of a negative community reaction to an episode labeled as psychogenic, saying, “Physicians and others are understandably reluctant to announce that an outbreak of illness is psychogenic because of the shame and anger that the diagnosis tends to elicit.”

(#litres_trial_promo)

An Unbearable Sweetness

Outbreaks of epidemic hysteria are not restricted to children and schools. They have been documented in adults too. One systematic review of cases of epidemic hysteria identified seventy outbreaks that occurred between 1973 and 1993 and found that 50 percent of them took place in schools, 40 percent in small towns and factories, and only 10 percent in other settings.

(#litres_trial_promo) The outbreaks usually involved at least thirty people, and often hundreds. Most outbreaks lasted less than two weeks, but 20 percent lasted more than a month.

One of the more improbable examples was the case of the “phantom anesthetist of Mattoon.” In 1944, over a period of a few weeks during the climax of World War II, many adult residents of Mattoon, Illinois, became convinced that an “evil genius” was on the loose in their town of fifteen thousand people. This unseen person would open bedroom windows and spray victims with a “sweet-smelling” anesthetic gas that would temporarily paralyze them but, strangely, leave others in the same room unaffected. Citizens banded together to form armed patrols, but the anesthetist was never caught. The local sheriff, fearing that an innocent person might be shot, eventually ordered the posses to disband. As one investigator of this outbreak dryly noted, “The ‘gasser’ hypothesis asserts that the symptoms were produced by a gas which was sprayed on the victims by some ingenious fiend who has been able to elude the police. This explanation…is widely believed in Mattoon at present. The alternative hypothesis is that the symptoms were due to hysteria.”

(#litres_trial_promo)

Another, more recent case occurred in 1990 among the Triborough Bridge toll employees in New York City. On February 16, workers began to complain of headaches, abdominal discomfort, dizziness, and throat and chest pain. More and more workers came down with the same symptoms over the next several days, with some of the ill workers noting what they described as a “sweetness” in the air. Symptoms were reported when workers were inside or near a toll booth, but they would subside soon after workers left the booths. The outbreak ended on February 22, when some of the workers’ superiors sat with them at the tolls. By that time, thirty-four workers had become ill enough to go to the hospital, and many others shared their symptoms. After spending hundreds of thousands of dollars searching in vain among dozens of potential culprits for a physical cause of the symptoms, it became clear to many that the illness was psychogenic. It forced 44 percent of the female workers to go to the hospital, almost twice the proportion of male workers with debilitating symptoms.

These cases share many characteristics of MPI. The symptoms tend to pop up in and spread through highly connected communities (with high network transitivity). These communities tend to be isolated and stressed. A physical culprit is seldom found. In most cases, the majority of those affected are women. It is not clear why the incidence in women and girls is higher, but it is possible that because women are inclined to discuss their symptoms, more sympathy cases result in other women. The fact that women have a more sensitive sense of smell might also play a role.

For some reason that is not well understood, smells, both real and imagined, are frequent triggers of modern outbreaks of MPI. This may have to do with the well-established connection between olfaction and emotions. Experiments have demonstrated that smell and emotion are both regulated by a part of the brain called the orbitofrontal cortex.

(#litres_trial_promo) Experiments have also shown that memories evoked by smell induce stronger emotions than those evoked by verbal descriptions of the same odor.

(#litres_trial_promo) Words are powerful, but one familiar whiff can jolt the mind into the past with more emotional intensity than can a signal from any other sense. This is called the Proust phenomenon, after the author who described a poignant memory inspired by the scent of a cookie. Smelling a perfume associated with a happy memory leads to more activity in the amygdala (a part of the brain involved in emotion and emotional memory) than seeing the bottle that the perfume comes in.

(#litres_trial_promo)

Paradoxically, the presence of official personnel—whether police officers, rescue workers, scientific investigators, or government officials—often worsens the epidemic, for it reinforces the belief that something serious is going on and that the situation is potentially dangerous. When these same officials attempt to provide reassurance that the situation is safe and that no cause was found, it typically generates deep suspicions among the emotionally charged populace that a cover-up is under way, especially because the official response was previously so substantial. Paranoia can spread too, undermining the very authority that is needed to bring an end to such an episode.

The recommended treatment for MPI outbreaks focuses on social networks and recognizes that social ties are the medium for spread. The psychological guidelines for emergency workers include “providing reassurance…using a calm and authoritative approach” and “separating those who are ill from those who are not.”

(#litres_trial_promo) As one expert put it, “You can only stop these things by being honest…. I could get caught up in this kind of thing too, as a parent or just a person. We all could. It’s a very powerful thing, and it needs to be respected and understood. And health officials shouldn’t be so scared to call a spade a spade.”

(#litres_trial_promo)

It’s often difficult to establish why exactly these epidemics start. Just as an unfamiliar noise can trigger a cattle herd to start running, many triggers can cause emotional stampedes. However, it is usually fairly simple to identify the initial cases. For example, in the African laughing epidemic, even though the investigators could not explain why it started, they easily located the first girls to have symptoms.

It only took a few people to start La Ola in the stadium in Mexico City or to get passersby to stop and look up at a window in New York City, and the same is true of MPI outbreaks. When a small group of people begin acting in concert or experiencing similar, visible symptoms, the epidemic can spread along social-network ties via emotional contagion, and large groups can very quickly become emotionally synchronized.

The present obsession with nut allergies in the United States may be a case in point. The number of schools declaring themselves to be entirely “nut free” is by all accounts rising. Nuts and staples like peanut butter are prohibited from campus, and so are homemade baked goods and any foods without detailed ingredient labels. School entrances have signs admonishing visitors to wash their hands before entering to safeguard students from possible contamination.

Approximately 3.3 million Americans are allergic to nuts, and even more, 6.9 million, are allergic to seafood. However, all told, serious allergic reactions to foods cause just two thousand hospitalizations per year (out of more than thirty million hospitalizations nationwide). And, at most, only 150 people (both children and adults) die each year from food allergies. Compare that to the fifty people who die each year from bee stings, the hundred who die from lightning strikes, and the forty-five thousand who die from motor vehicle accidents. Or compare that to the ten thousand children who are hospitalized each year for traumatic brain injuries acquired during sports, or the two thousand who drown, or the roughly thirteen hundred who die from gun accidents. Yet there are no calls to end athletics. There are likely thousands of parents who rid their cupboards of peanut butter but not guns. And more children assuredly die walking or being driven to school each year than die of nut allergies.

The question is not whether nut allergies exist, or whether they can occasionally be serious, or whether reasonable accommodations should be made for the few children who have documented serious allergies. The question is, what accounts for society’s extreme response to nut allergies? Not surprisingly, the response bears many of the hallmarks of MPI. A few people have clinically documented concerns, but others who do not then copy the behaviors of those who do. Anxiety spreads from person to person to person, and a sense of proportion and the ability to be reassured are lost.

Well-intentioned efforts to reduce nut exposure actually fan the flames since they indicate to parents that nuts are a clear and present danger. This encourages more parents to worry, which fuels the epidemic. It also encourages more parents to have their kids tested, thus detecting mild and meaningless allergies to nuts. And, finally, this encourages still more avoidance of nuts, which may actually lead to a rise in true nut allergies because lack of exposure to allergens early in life is thought to contribute to the onset of allergies later.

(#litres_trial_promo)

MPI is a pathological phenomenon, but it takes advantage of a nonpathological process that is fundamental in humans, namely, the tendency to mimic the emotional state of others. Real laughter also can be contagious and so can real happiness. But comparing epidemic hysteria to these more normal processes is like comparing the stampede of a herd to its more usual and orderly migration.

Tracking the Spread of Emotions

Measuring the subjective experience of emotions (as compared with their visible, biological, or neurological manifestations) requires asking people how they are feeling. One of the more systematic ways of doing this is known as the experience-sampling method. This method uses a series of alerts (such as signals sent to a beeper or cell phone) at unexpected times to prompt subjects to document their feelings, thoughts, and actions while they are experiencing them.

(#litres_trial_promo) The result is a thorough picture of the ups and downs of subjects’ daily emotional lives.

One of the advantages of this method is that it allows groups of interacting people to be evaluated simultaneously in real time. For example, one team of investigators, interested in the spread of emotions within families, outfitted fifty-five families (consisting of a mother, father, and one adolescent) with beepers for one week. The participants were beeped roughly every 90 to 120 minutes between 7:30 a.m. and 9:30 p.m., and a total of 7,100 time points were observed in these 165 individuals. Various emotional states were measured, such as whether the subjects were happy or unhappy. Although the investigators could not rule out the possibility that the entire family was simultaneously exposed to one thing that made them all sad or happy at once (a confounding effect that we will discuss in greater detail in chapter 4 (#litres_trial_promo)), they did try to tease out how emotions spread within these families.

The strongest path was from daughters to both parents, while, conversely, the parents’ emotional state appeared to have no effect on their daughters. Fathers’ emotions affected their wives and their sons but not their daughters. This appeared to be especially true when fathers returned from work: when dad came home in a lousy mood, he soon made the whole household miserable.

(#litres_trial_promo)

A similar method has been used to examine the transmission of emotions among teams of nurses, athletes, and even accountants.

(#litres_trial_promo) In such professional settings, a key question was whether one fired-up team member could improve the mood and thus the performance of his teammates. Not surprisingly, positive mood is associated with a range of team-performance-enhancing changes, including greater altruistic behavior, increased creativity, and more efficient decision making. A nice demonstration involved outfitting thirty-three professional male cricket players with pocket computers that recorded their moods four times a day during a match (which can have the insane duration of five days). There was a strong association between a player’s own happiness and the happiness of his teammates, independent of the state of the game; further, when a player’s teammates were happier, the team’s performance improved.

The Spread of Happiness

Despite the biological and psychological evidence for emotional mimicry, and the numerous cases of MPI arising from epidemic anxiety, until recently little was known about the precise role of social networks in the spread of emotions. Yet, the MPI cases suggest that emotions spread far and wide, flowing through social-network ties from person to person to person, and that there should be a normal analogue to this pathological phenomenon. Indeed, there can be waves of emotions in the vast fabric of human social relationships, so that people in particular locations in the social network have one emotional experience, and others elsewhere who come under different influences have a different experience altogether.

Strangely, while researchers in diverse fields, including medicine, economics, psychology, neuroscience, and evolutionary biology, have identified a broad range of stimuli of individual human happiness, they have not addressed a key (perhaps the key) determinant: the happiness of others. It may be obvious that our friends and family can make us happy, but before we undertook our own investigation, no one had ever explored how happiness can spread through social networks from person to person to person.

We became curious about this. We were particularly interested in determining whether the spread of emotions occurred not just between you and your friends (dyadic spread) but also between you and your friends’ friends, and their friends, and beyond (hyper-dyadic spread). How far did emotions travel in the network? And were there geographic or temporal constraints on the spread?

Our first step in answering these questions was to assemble a data set that had measures of emotions and social connections over time. (We discuss that process in chapter 4 (#litres_trial_promo).) We then created a graph of the social network of happiness, as shown in plate 1. This illustration shows ties among siblings, friends, and spouses in a sample drawn from 12,067 people originally from Framingham, Massachusetts, in the year 2000, along with their levels of happiness. No one had ever plotted such a graph before. One thousand twenty people are represented, and each node is colored on a spectrum from blue (unhappy) to yellow (happy) according to the subject’s level of happiness. Looking at this image suggests two observations. First, unhappy people cluster with unhappy people in the network, and happy people cluster with happy people. Second, unhappy people seem more peripheral: they are much more likely to appear at the end of a chain of social relationships or at the edge of the network.

(#litres_trial_promo)

Clustering of this kind in social networks can arise from a variety of processes. Happy people might choose each other as friends or be exposed to the same environments that cause them all to be happy at the same time. But our analyses allowed us to adjust for these effects. And we found that clustering is also due to the causal effect of one person’s happiness on another’s. Mathematical analyses of the network suggest that a person is about 15 percent more likely to be happy if a directly connected person (at one degree of separation) is happy. And the spread of happiness doesn’t stop there. The happiness effect for people at two degrees of separation (the friend of a friend) is 10 percent, and for people at three degrees of separation (the friend of a friend of a friend), it is about 6 percent. At four degrees of separation, the effect peters out. Here we have our first evidence of the Three Degrees of Influence Rule. Emotions (and, as we will see later, norms and behaviors) spread in social networks from person to person to person, but they do not spread to everyone. Just as a ripple in a pond eventually fades away, so too does the ripple of an individual’s happiness fade through the social network.

At first glance, these effects may not seem very significant. But compare them to the effect of having a higher income. An extra $5,000 in 1984 dollars (which corresponds to about $10,000 in 2009 dollars) was associated with only a 2 percent increased chance of a person being happy. So, having happy friends and relatives appears to be a more effective predictor of happiness than earning more money. And the amazing thing is that even people who are three degrees removed from you, whom you may have never met, can have a stronger impact on your personal happiness than a wad of hundreds in your pocket. Being in a particular spot in a social network, exposed to people with particular feelings, has important implications for your life.

It is well known that having more friends and relatives is much more likely to put a smile on your face than having more cash.

(#litres_trial_promo) But past research had never considered why friends matter so much. There are at least two possibilities. First, the existence of the social relationship itself may improve your happiness—this is a structural effect of the network on you (the second rule of social networks described in chapter 1 (#u3bfc2f77-1cce-5f27-adf5-ae189eb589b0)). As we discuss in chapter 7 (#litres_trial_promo), we are hardwired to seek out social relationships, so it is not surprising that we feel pleasure or reward when we spend time with friends and family. Second, friends and relatives make us susceptible to emotional contagion, so our friends’ emotional states affect our own (the third rule of social networks).

While both of these mechanisms probably contribute to people’s happiness, our evidence suggests that contagion may be the more important of the two. We found that each happy friend a person has increases that person’s probability of being happy by about 9 percent. Each unhappy friend decreases it by 7 percent. So if you were simply playing the averages, and you didn’t know anything about the emotional state of a new person you just met, you would probably want to be friends with her. She might make you unhappy, but there is a better chance she will make you happy. This helps to explain why past researchers have found an association between happiness and the number of friends and family. But once we control for the emotional states in one’s friends, we find that having more friends is not enough—having more happy friends is the key to our own emotional well-being.

This does not mean that the structure of the social network is unimportant. Amazingly, it is not just the number of dyadic ties that has an impact; the number of hyperdyadic ties also influences a person’s happiness. When we measured the centrality of each person in the social network, we found that people with more friends of friends were also more likely to be happy. And, more remarkably, this was true even among people who had the same number of direct social relationships. This means that the more friends your friends have (regardless of their emotional state), the more likely you are to be happy.

One might wonder if there is a chicken-and-egg problem here. After all, it is possible to imagine that when we become happier, we tend to attract more friends, and more friends who have lots of friends. This would mean that happiness is driving the network rather than the other way around. But when we examined how the network changes over time, we found that happy people do not tend to become more central. So having a wide social circle can make you happy, but being happy does not necessarily widen your social circle. Being located in the middle of the network leads to happiness rather than the other way around. The structure of your network and your location in it matter.

Given how important direct interaction seems to be for emotional contagion to occur, we also theorized that the effect of the happiness of your social contacts on your emotional state should depend on how near or far they are. The idea is that people who live nearby are more likely to be in contact and therefore more likely to pick up on each others’ moods. Geographic distance can be used as a proxy for frequency of social interaction. In our study, about one in three people live within a mile of their closest friend, but there is a lot of variation, and some friends live thousands of miles apart. We found that when a friend who lives less than a mile away becomes happy, it can increase the probability that you are happy by 25 percent. In contrast, the happiness of a friend who lives more than a mile away has no effect. Similarly, if your spouse lives with you and he or she becomes happy, then your probability of happiness goes up, but spouses who do not live together (because they are separated) have no effect on each other. A happy sibling who lives less than a mile away increases your chance of happiness by 14 percent, but more distant siblings have no significant effect. And happy next-door neighbors also increase your chance for happiness, while neighbors who live farther away (even on the same block) have no significant effect.

All these findings suggest the importance of proximity among people whose emotions influence each other, and the impact of immediate neighbors suggests that the spread of happiness may depend as much on frequent face-to-face interaction as on deep personal connections. While in this case we are considering the spread of a dis-positional state of some duration, these findings are also in keeping with the work on facial mimicry we discussed earlier.