Contagion theory is one of the earliest theories developed to study collective action

2 Epidemic Modeling of the Onset of Social Activities

Additional research on teenage sexual behavior has borrowed from the epidemiology literature by framing teenage sexual behavior and pregnancy as an epidemic. The epidemic framework extends the idea of disease transmission to social behaviors—over time, behaviors spread among networks of peers. Rodgers and Rowe (1993) introduced the social contagion theory of adolescent sexual behavior as a probabilistic model in which transmission of the behavior (i.e., sexual intercourse) depends on features of adolescent dyads, including adolescents’ status with regards to the behavior (i.e., virgin) and their pubertal development (Rodgers, 1996). Transition rates can vary by gender and for different types of behaviors (Rodgers, Rowe, & Buster, 1998). This model conceptualizes adolescent sexual behavior as a marker of transition between adolescence and adulthood that encompasses a variety of attitudes and behaviors that develop gradually, rather than as a single act (Rodgers, 1996). The spread of behavior occurs through interacting social networks of adolescents, such as peers and family members, especially siblings. Social influence passes from the adolescent who has performed the behavior to those adolescents who have not performed the behavior. Initially, few adolescents have engaged in sexual behavior, but, over time, sexual behavior spreads, and the spread occurs in tandem with other relevant transition behaviors, such as delinquency, drinking, and smoking (Rodgers, 2003; Rodgers et al., 1998). Attitudes and decisions about sexual behavior and the consequences of sexual behavior are therefore shaped within the context of peer relationships.

Rodgers, Rowe and Buster (1998) used a nonlinear Epidemic Modeling of the Onset of Social Activities (EMOSA) model to reflect the process by which teenage girls get pregnant and the relationship between sexual behavior and pregnancy. The estimated pregnancy probabilities for girls of different ages and different virginity statuses are similar across both age and time-since-virginity, suggesting that younger and older girls are equally likely to get pregnant regardless of how much time has passed since becoming sexually active. These results suggest a model in which sexual behavior is transmitted from adolescents who have had sex to those who have not, and as this transmission occurs, some girls become pregnant (Rodgers et al., 1998). Again, although social contagion theory can help us understand the spread of teenage sexual behaviors and pregnancy, there may be different factors and processes involved in teenagers’ decisions to actually give birth to a child after becoming pregnant. The EMOSA model would need to be extended to predict the probabilities of teenage childbirth to determine whether age, time-since-virginity, and other factors (e.g., family background characteristics) differentially influence the probability of teenage childbirth.

Social Factors

One of the puzzles about BPD is why it was first described only 80 years ago. Madness and melancholia have been described throughout recorded history. Is it possible that this disorder may not have existed prior to that time? If that were the case, BPD would not be the only example. Eating disorders were rare until the 19th century, and bulimia nervosa only emerged as a frequent problem during my own lifetime as a psychiatrist.

A likely explanation is that some of the behaviors associated with BPD, such as cutting and overdosing, are due to social contagion, in which young people show distress in ways that are transmitted through peer groups and the media. For example, emotion dysregulation could have been expressed in the past through somatic symptoms, but in a modern society is more likely to be associated with impulsive behaviors.

This is a difficult area to research. But I have long been interested in the idea that BPD is a product of modernity (Paris, 1996, 2013). In the past, most people lived in small villages, had the same jobs as their parents, and chose a spouse from a narrow social network. We live in an age where people have more choices. But not everyone finds it easy to do what modern society expects, such as establishing a unique identity.

Several lines of research support this hypothesis. One is historical: the absence of records describing BPD patients in past times. Another is the likelihood that BPD symptoms are on the increase. The other is crosscultural—evidence that BPD is less common in developing countries, but is now being recognized in large urban areas that are becoming modernized. Moreover, psychological risk factors will be more pathogenic in problematic social environments, where the road to resilience is often blocked.

Collective behavior

Early North American theorizing about collective action (a particular focus was behavior in crowds but also included social movements) referred to this set of phenomena as collective behavior. Explanations for collective behavior that were developed during this period drew heavily upon the imagery of the community-lost thesis. Collective behavior was said to occur when a large number of people failed to accept some of the dominant values, norms, and/or leaders in a society. One of the earliest North American sociologists to give this problem serious attention was Herbert Blumer. Blumer’s work followed in the footsteps of earlier European theorists (such as Le Bon and Tardé) in that he concurred that ‘social contagion’ was the main process underlying collective behavior. However, he contrasted circular reaction with ‘interpretive’ behavior. Interpretive behavior refers to actions that are guided by thoughtfully interpreting the behavior of others.

Also working within the interpretive tradition (e.g., the symbolic interaction perspective), and hence, building upon Blumer’s work, Turner and Killian developed ‘emergent norm theory’ to account for collective behavior. In contrast to Blumer, Turner and Killian argued that there is substantial diversity among the participants in collective behavior. The emergent norm concept is used to explain the empirical fact that crowds are not unanimous in their feelings and expressions; individuals are quite diverse in terms of their sentiments as well as motives and actions. An illusion of unanimity arises or is constructed because the behavior of part of the crowd becomes salient and is perceived both by observers and crowd members as being the sentiment of the whole crowd. Variant views and divergent behavior are overlooked. Shared definitions about crowd behavior becomes an emergent norm. This norm inhibits behavior contrary to it and leads participants to implement sanctions against norm violators.

5.24.6.5 Cultural Factors

A complete understanding of eating disorders requires attention to the cultural forces selectively impinging upon on young girls and women in modern times. One of the most pernicious has been the intense pressure to diet and to engage in strict weight control to meet unrealistic standards for thinness (Garner and Garfinkel, 1980; Garner et al., 1980; Stice et al., 2017; Thompson et al., 1999; Vandereycken and Hartley, 1996). Young women today are totally immersed in the cultural admiration of a physical form for women that has little to do with the actual shape of most women in our society. The disconcerting result is a norm in which women report being dissatisfied with their shapes and to feeling guilty about eating even reasonable amounts of food. It has been increasingly recognized that dieting can play a direct role in causing a range of symptoms such as binge-eating and mood. The values surrounding slenderness have become sufficiently embedded in our cultural value system so that many of the symptoms required for eating disorders are not viewed as unusual or abnormal by members of the general public (Huon et al., 1988). There is even some evidence that eating disorders may have developed a positive social stereotype and, in some instances, may be spread by social contagion (Vandereycken, 2011). Recognition of the impact of cultural factors on norms related to dieting and weight control has led to the conclusion that eating disorders may develop in those without underlying personality disturbances or family dysfunction, although secondary disruption in both areas may be present by the time the person presents for an assessment (Garner, 1997). However, much of the current conceptualization of the etiology of the thin-ideal internalization suggests a significant effect of the shared environment since most women are exposed to high levels of thin idealization in the media and many women also are exposed to parallel values from parents and peers. In a longitudinal study of early adolescent girls from fifth through eighth grade, one study reported that mothers, fathers, and peers each played an important role in the development of bulimic symptoms by affecting their body dissatisfaction, which was related to later dieting behaviors, depressive symptoms, and bulimic symptoms (Salafia and Gondoli, 2011). Moreover, results from this study suggested that peers were a stronger influence than mothers and fathers. In addition to considering exposure to risk factors, it is also important to understand the role of protective factors in the expression of eating disorders. One study found that women who had a high level of self-determination, where behaviors are experienced as reflective of autonomous choice, were less influenced by “thin ideal” media exposure than those with low self-determination (Mask and Blanchard, 2011).

It is also possible that genetic factors contribute to the internalization of “thin ideals” with some being more affected and others relatively immune to the cultural messages. In an attempt tease out the genetic and environmental influences, Suisman et al. (2012) examined “thin ideal internalization” in participants from a twin registry. They reported significant genetic and nonshared environmental influences suggesting a strong genetic predisposition to thin-ideal internalization that parallels heritability of disordered eating and eating disorders (Bulik et al., 2010; Sullivan et al., 1998).

Cultural factors directly contribute to eating disorders and/or the influence may be mediated by psychological vulnerability. Nevertheless, it is important to emphasize that psychological theorizing related to eating disorders must account for cultural factors as well as the susceptibility of some to the influence of these factors in the development of eating disorders.

The Principle of Phylogenetic Continuity

One clue that may provide a link and serve as an organizational principle among the various components of social cognition, such as empathy, theory of mind, imitation, mimicry, and related processes is their phylogenetically continuous characteristic (Adolphs, 2003a). Frans de Waal and Stephanie Preston have argued persuasively that the roots of empathy can be traced to our closest living relatives, the chimpanzees. Indeed, a number of studies have now shown that humans and other mammals (mice, dogs, birds, monkeys) respond empathically to a conspecific they have encountered previously, and thus to the perceived overlap (e.g., familiarity or similarity) between self and other (Preston & de Waal, 2002; Silva, Bessa, & de Sousa, 2012). (Jones and Josephs 2006) found that dogs react to their owner’s stress with an increase in negative emotional arousal. Similarly, (Plotnik and de Waal 2014) showed that following a distressing event, elephants affiliated significantly more with other conspecifics through directed physical contact and vocal communication.

It can be argued that successful integration into a social group requires the need for not only effective but also affective communication, something that characterizes a variety of animals (Preston & de Waal, 2002). In environments with ever increasing social complexity, as seen in many large primate groups, it becomes important to be able to communicate information as well as the emotional context of that information with other members of the group. Empathy appears to enhance social function by tuning our response to the emotions of another. Hence, knowing the emotional state of the other, and making decisions based on that understanding, facilitates the most advantageous response to the situation even when fleeing or joining in a fight.

The more fundamental abilities from which empathy might have arisen include social contagion, response facilitation, and emulation. All of these are considered categories of social learning—from observational to active learning. Some of the most salient emotional responses are acquired through observing the behaviors of others, especially those involved in avoidance and withdrawal, such as taste aversion and predator fear (Tomasello, Carpenter, Call, Behne, & Moll, 2005). Thus, from an evolutionary perspective, empathy has survival value in that it helps individuals gather and hunt for food, detect predators, enhance courtship, and ensure reproductive success. It can be considered one of the most important social skills in humans because it not only provides for strong relationships but also affects social learning as well. The problem with making such an evolutionary argument, however, is that it is difficult to see how natural selection works at the level of superordinate categories, such as empathy. It is more logical to suggest that it is the more primitive functions that are selected and that those in turn are combined to produce complex responses. But, just what are these constitutive components of empathy?

To a large degree, our worldview is consistent with that of (Decety and Jackson 2004), who argue that the basic macro components of empathy are mediated by specific neural systems. Furthermore, they have suggested that these include shared neural representations, self-awareness, mental flexibility, and emotion regulation. We agree that mirroring processes, for example, by which we also mean the shared neural representation of information about self and other, along multiple levels and across multiple modalities of the CNS, are critical components. We believe that natural selection for these processes produced the abilities to mimic and imitate, which in turn generated the ability to learn through such mechanisms. Learning may have facilitated the evolution of the abilities to resonate and respond to the emotions of others (empathy) as well as to represent the beliefs and intentions in other minds (a theory of mind). Empathy and theory of mind, plus the underlying functional infrastructure, likely allowed for social learning and the faithful transmission of culture (Gintis, 2003; McElreath, 2003).

Within the context of the H–H model, certain social cognitive skills—notably the ability to represent another conspecific’s mind or theory of mind—distinguish humans and perhaps apes from other animals. The implication is that the elaboration of social cognition in different animals (e.g., nonhuman primates, dogs, rodents) may depend on neural resources, with larger brains and larger memory capacities allowing for the development of more complex processes. It is widely agreed that compared to other species, humans exhibit more advanced and flexible forms of social behaviors, including empathy, self-awareness, and perspective-taking (Decety & Svetlova, 2012).

2.16.7 Retrieval Errors and Other Retrieval Phenomena

Earlier, we reviewed research that shows that repeated retrieval improves memory. As Bjork (1975) noted, retrieval is a memory modifier. Interestingly, the act of retrieval can also reduce memory accuracy. People recall emotional, significant, or entertaining events from their lives often, and it is all too common to embellish the events from one telling to the next or from one audience to another. Bartlett’s (1932) classic study is often cited to illustrate how memory output can change from one recall to the next. In this study, people were asked to read a Native American story called ‘The War of the Ghosts.’ Importantly, the significance of many details in this story was not apparent to people of different cultural backgrounds. The more times that people were asked to retell the story, the more the stories became distorted such that subjects omitted unfamiliar details and inserted materials to make the story consistent with their schemas. These dramatic changes across repeated retrievals have since been replicated in a study by Bergman and Roediger (1999).

Bartlett’s (1932) study and Bergman and Roediger’s (1999) critical replication show that individuals modify story output even without any intervention from outside sources. It is easy to imagine then that the social situations individuals encounter can change the contents of what individuals might retrieve from one occasion to the next. For example, social situations often dictate whether a story should be told in an accurate or entertaining fashion. In a study by Dudukovic et al. (2004) participants were asked to either retell a story with a goal to be accurate, or with a goal to be entertaining. While the participants did not differ on a later recognition test, they did differ on a later recall test. The participants who had originally told the story accurately recalled more information with less exaggeration than the participants who had originally told the story for entertainment. Thus, the way that we recount information to others influences the way that information is later recalled. As we will describe next, retrieval errors can also creep into individual performance when others provide input during retrieval.

In a previous section on retrieval in a social context, we discussed evidence that shows positive influences of input from others. But individuals also make more retrieval errors under certain conditions if they previously received erroneous input from others. In a study that assessed the effects of social contagion, Roediger et al. (2001) presented subjects with everyday scenes (e.g., a kitchen) during the study phase. Later, subjects recalled the scenes along with a confederate who inserted related but nonstudied items during recall (e.g., toaster). On a final test where subjects engaged in recall alone, they falsely recalled related but nonstudied items more often if they had been inserted by a confederate in the earlier recall phase than if no mention of them had been made.

Basden et al. (2002) reported similar effects of social input in a study where they presented semantically related ‘DRM’ (Deese-Roediger-McDermott) lists during study and constructed a perceived group-recall situation with the use of interconnected computers. DRM lists consist of thematically related words such as ‘dream,’ ‘bed,’ ‘night,’ etc., where a critical word such as ‘sleep’ is missing. In individual memory studies, subjects erroneously recall these critical nonpresented words at levels as high as true recall and also give ‘remember’ responses indicating vivid memory for having seen them before (Roediger and McDermott, 1995). In Basden et al.’s study, subjects engaged in perceived group recall followed by individual recall. During perceived group recall, the subjects were led to believe that they viewed the responses of other group members on the computer screen during recall but in fact the generated responses were controlled by the experimenter. In one condition of perceived group recall, subjects saw the critical, nonpresented lures, and in another condition these items were not included in the supposed responses from other members. Subjects included more erroneous responses in their final individual recall protocols if they had previously participated in one of the two perceived group-recall phases than if they had not participated in the perceived group-recall phase. In this way, individual retrieval can be socially influenced. The process of collaboration can lead to individual retrieval benefits as discussed in the previous section but also to retrieval errors, as these studies show (also see Basden et al., 2000, 2002; Reysen, 2005, for related findings).

Another topic of considerable interest in memory retrieval focuses on the subjects’ ability to identify the source of information they recall – a phenomenon called reality monitoring (Johnson and Raye, 1981) The general approach here is to ask subjects whether the item they recalled (or recognized) was presented to them (i.e., the item originated from perception) or was internally generated (i.e., the item was something they imagined or dreamed). According to this framework (e.g., Johnson, 1991; Johnson et al., 1993), people do not explicitly tag memories with source information. Rather, they typically make source attributions based on a generalized evaluation of whether a memory’s qualities match expectations. These judgments capitalize on the average differences of the characteristic qualities of memories from different sources. For instance, perceived events tend to include more information about perceptual, temporal, spatial, and affective characteristics and less information about cognitive processes than imagined events. A judgment of ‘perceived’ rather than ‘imagined’ should therefore be given if the evaluation of a memory’s qualities results in a great deal of information about perceptual and spatial details, accompanied by little information about the cognitive processes that took place during encoding. Attributing a memory to the source for which that memory’s qualities are most characteristic maximizes the odds of accurately judging the memory’s source. Reality-monitoring failures occur when people falsely claim either that something was perceived when it was actually internally generated or that something was internally generated when it was actually perceived. A detailed review of this topic can be found in a chapter on source monitoring by S. Lindsay (see Chapter 2.19) in this volume.

These processes in reality monitoring constitute yet another form of retrieval, one that is characterized by metamemory judgments, because subjects make judgments about information retrieved from memory. In an earlier section on retrieval mode, we described remember and know judgments, which can also be considered metamemory judgments because subjects report the quality of memory for the information they retrieve (see Rajaram and Roediger, 1997; Rajaram, 1999; Roediger et al., 2007). There are also other well-known metamemory judgments such as feelings of knowing (see Koriat, 1995) and the tip-of-the-tongue state (see Brown, 1991; Schwartz et al., 2000) that researchers study to find out subjects’ sense of what they can retrieve even when recall does not succeed. These judgments reveal interesting – metacognitive – aspects of the retrieval process as subjects make judgments about the likelihood of retrieval under certain circumstances. In the feeling-of-knowing state (see Koriat, 1995) subjects can reliably report whether they can recognize an item on a multiple-choice test even though they were unable to recall that item, and in the tip-of-the-tongue state, people can reliably indicate whether or not the information they are trying to retrieve is on the tip of their tongue and could be retrieved. We recommend chapters by A. Koriat on control processes in remembering (see Chapter 2.18) and by A. S. Brown on the tip-of-the-tongue states (see Chapter 2.22) for detailed discussions of these topics.

In conclusion, the ways in which retrieval conditions are arranged to a large extent determine how much memory accessibility can improve. But in many situations, retrieval can also act as a memory modifier and can do so in systematic ways. Such situations can lead to systematic errors in retrieval, as revealed by the DRM effect. Furthermore, when recalling information, people are motivated not only to present a coherent story, but also to tell the story with a particular purpose (for example, to be entertaining). Both of these motivations can serve to lower overall memory accuracy. Retrieval errors can also be the result of social influences (believing that you saw something that someone else endorsed seeing), or reality monitoring errors (believing that you saw something that you only imagined seeing). Finally, many meta-memory processes such as the tip-of-the-tongue phenomenon also modulate the success of retrieval.

3.2.2 Social contagion

Fig. 4b shows mean false alarms for items that were initially unstudied by each subject, but studied by one if not both other group member(s). A 2 × 3 ANOVA with the factors of item type (weak contagion items, strong contagion items) and retrieval condition (II, CI, ICI) showed a significant main effect of retrieval condition, F(2,87)=15.01, MSE=0.06, p<.001, ηp2=0.26, a significant main effect of item type, F(1,87)=16.83, MSE=0.02, p<.001, ηp2=0.16, but no significant interaction of the two factors, F(2,87)=2.09, MSE=0.02, p=.130, ηp2=0.05. To clarify this pattern, we again ran further analyses.

An ANOVA contrasting the individual and collaborative retrieval conditions (II vs. CI) showed a significant main effect of retrieval condition, F(1,58)=30.93, MSE=0.06, p<.001, ηp2=0.35, a significant main effect of item type, F(1,58)=10.90, MSE=0.02, p=.002, ηp2=0.16, and a significant interaction of the two factors, F(1,58)=4.63, MSE=0.02, p=.036, ηp2=0.07. Follow-up tests on false alarms in the individual retrieval condition alone again showed no differences between strong and weak contagion items (29.7% vs. 27.0%), t(29)=1.03, p=.310, d=0.19. In the collaborative retrieval condition, however, strong contagion items evoked higher false alarm rates than weak contagion items (58.7% vs. 46.0%), t(29)=3.28, p=.003, d=0.60. False alarm rates for both item types were enhanced in the collaborative vs. the individual retrieval condition, but social contagion was again larger for strong (58.7% vs. 29.7%), t(58)=6.01, p<.001, d=1.55, than for weak contagion items (46.0% vs. 27.0%), t(58)=3.82, p<.001, d=0.99.

To examine if individual retrieval prior to collaboration influenced social contagion, we conducted another ANOVA contrasting the two collaboration conditions. The ANOVA revealed a significant main effect of item type, F(1,58)=16.21, MSE=0.03, p<.001, ηp2=0.22, and a significant main effect of retrieval condition, F(1,58)=6.91, MSE=0.06, p=.011, ηp2=0.11, but no significant interaction between the two factors, F(1,58)<1.00, MSE=0.03, p=.778, ηp2=0.001. False alarm rates were also higher for strong compared to weak items in the collaborative retrieval condition with a prior individual test (46.3% vs. 35.3%), t(29)=2.48, p=.019, d=0.45. Relative to the condition without prior individual retrieval, false alarm rates were however significantly reduced, and this was the case for both strong contagion items (46.3% vs. 58.7%), t(58)=2.22, p=.030, d=0.57, and weak contagion items, (35.3% vs. 46.0%), t(58)=2.15, p=.036, d=0.55.4

5 Future perspectives

Recently, as the critical amount of people is easily accessible via the Internet and the mass media, the likelihood and the severity of various health related social contagions increases. This fact was clearly shown by several recent reports [188–191]. Generally, the importance of complete health is strongly stressed by the mass media, which leads to a heightened consciousness of health issues, and also to an increased feeling of vulnerability [142,192]. Moreover, a new phenomenon called cyberchondria was described [193,194]. Individuals with cyberhondria show an excessive use of the Internet for health and illness related information [195], which in turn increases health anxiety and generates more search activity [196].

A variety of specific questionnaires has been developed to assess individual sensitivity to the aforementioned phenomena, such as the Modern Health Worries Scale[142], the Perceived Sensitivity to Medicines Scale[197], the Environmental Hypersensitivity Symptom Inventory[198], and the Cyberchondria Severity Scale[195]. These questionnaires (and the underlying constructs) focus on the specific content of the thought processes thus are more specific. In contrast, the SSAS assesses a more general and more evaluative-emotional characteristic. It is short and easy to administer; moreover, the underlying construct is less obvious, which means that the results will be less biased by demand characteristics and social desirability motive.

Although the explanatory power of the above described new approach to SSA seems to be higher than that of the original concept, the quality and quantity of supporting empirical findings are clearly not satisfying. To draw a final conclusion, well-designed empirical studies are needed. For example, experiments that compare the psycho-physiological reactions of individuals with high and low levels of SSA to various external and internal threats, or longitudinal studies testing the predictive power of SSA to the development of IEIs or other functional somatic syndromes. Also, systematic research is needed to shed more light on the relations among SSA, disgust, and generalized hypervigilance.