Abstract This research investigates whether a contextual factor—social density, defined as the number of people in a given area—influences consumers’ propensity to share information. We propose that high- (vs. low-) density settings make consumers experience a loss of perceived control, which in turn makes them more likely to engage in word of mouth to restore it. Six studies, conducted online as well as in laboratory and naturalistic settings, provide support for this hypothesis. We demonstrate that social density increases the likelihood of sharing information with others and that a person’s chronic need for control moderates this effect. Consistent with the proposed process, the effect of social density on information sharing is attenuated when participants have the opportunity to restore control before they engage in word of mouth. We also provide evidence that sharing information restores perceived control in high-density environments, and we disentangle the effect of social density from that of physical proximity. social density, compensatory control, word of mouth Word of mouth (hereafter WOM), defined as “informal communications directed at other consumers about the ownership, usage, or characteristics of particular goods and services, and their sellers” (Westbrook 1987, 261), has widely emerged as the main factor affecting consumers’ buying behavior in many categories, driving up to 50% of purchasing decisions (Bughin, Doogan, and Vetvik 2010). Recent technological and media trends have had a dramatic impact on WOM communication. Indeed, in today’s interconnected world, internet usage has been rapidly moving from desktop to mobile devices—so much so that now more than half of the world’s web traffic comes from mobile phones (Kemp 2017). This means that companies’ online shareable content is increasingly accessed from mobile devices while consumers are situated in different locations and contexts. Consumers might share that content online (e.g., via social media) or even more often offline (e.g., face-to-face with their friends; Keller and Libai 2009). Moreover, consumers are increasingly sharing information about their offline experiences while they are on the go by posting information on social media using their mobile devices (Shankar et al. 2010; You, Vadakkepatt, and Joshi 2015). In fact, it is extremely common to see consumers sending texts or emails while browsing stores or reading newspapers in coffee shops. In addition to WOM that occurs spontaneously, marketers themselves provide consumers with opportunities to comment online or offline. For example, events such as conferences and conventions try to increase audience participation and buzz by encouraging participants to talk about their experiences on social media, such as Twitter and Facebook, while the event takes place. Furthermore, leveraging the development of communication and geolocation technologies, companies spur online or offline WOM by targeting consumers with shareable content in specific locations and contingencies. For instance, companies can use indoor proximity systems that trigger actions on mobile devices, such as a check-in on social media when consumers enter a specific store or venue to signal consumers’ location to others. To sum up, consumers are reached by shareable content while they are located in social environments characterized by the presence of few or many other people (e.g., coffee shops, conferences, or subway trains). Thus, one relevant question is whether and how the presence of others might play a role in affecting information sharing. In this research, we investigate whether consumers’ general propensity to share information increases as the number of people in their environment increases. To clarify, our focal independent variable is a feature of the environment in which a consumer is situated: social density, defined as the number of people in a given area (typically indexed as the number of people per unit area; Andrews et al. 2015; Baumeister and Vohs 2007; Graumann and Moscovici 1986; Oldham and Rotchford 1983). Of note, we use the term density as synonymous to social density henceforth (e.g., a high-density environment is an environment characterized by high levels of social density—i.e., a large number of people). Our focal dependent variable is WOM. WOM typically includes face-to-face discussions, as well as online information sharing (e.g., online conversations, mentions, and reviews; Berger 2014). The specific focus of this research is the likelihood that WOM occurs. Specifically, drawing on research on social density (Rodin 1976; Sherrod and Cohen 1978), the psychological drivers of WOM behavior (Berger 2014; Hennig-Thurau et al. 2004; Sundaram, Mitra, and Webster 1998), and compensatory control (Cutright 2012; Cutright, Bettman, and Fitzsimons 2013; Kay et al. 2010; Langer 1975), we propose that high- (vs. low-) density environments (i.e., physical environments in which there are more vs. fewer people) make consumers experience a decrease in their feeling of personal control. This, in turn, makes them more likely to engage in information sharing in order to restore their lost sense of control. We present six studies that test and ultimately support this prediction. This research contributes to the extant literatures on compensatory control (Cutright 2012; Cutright et al. 2013; Kay et al. 2008; Kay et al. 2009, Kay et al. 2010; Langer 1975) and the psychological motives underlying WOM behavior (Berger 2014) by proposing an underexplored contextual factor—social density—as one condition under which individuals seek to restore their sense of control, and by proposing WOM as a means to restore control. We also contribute to extant knowledge about the role of social density in marketing (Andrews et al. 2015) by testing its effect on WOM behavior. Lastly, this research has significant practical implications. Companies might increase the effectiveness of their communication activity by sending targeted information to consumers who are in high-density environments, because these consumers might be particularly inclined to share this information either online or offline. THEORETICAL BACKGROUND Abundant research has investigated the underlying psychological processes that explain why individuals share information (see Berger 2014 for a review). Prior work has mostly focused on understanding what type of information individuals are more likely to share—for example, positive versus negative WOM (Godes and Mayzlin 2004), arousing versus nonarousing content (Berger and Milkman 2012), and self-presentational versus useful content (Barasch and Berger 2014). This work has also demonstrated that individuals consider WOM communications to be instrumental in fulfilling certain psychological needs (e.g., self-enhancement, De Angelis et al. 2012; social status, Gatignon and Robertson 1986; emotion regulation, Berger and Milkman 2012). However, amidst this research on the content characteristics and psychological drivers of WOM communications, little is known about how the environment itself affects consumers’ information-sharing behavior. Recent research in marketing has shown that social density positively impacts consumers’ receptivity to mobile advertising (Andrews et al. 2015). However, to the best of our knowledge, no studies have shed light on how social density specifically affects WOM behavior. This research is thus the first to propose that social density has a positive impact on consumers’ likelihood of engaging in WOM behavior, an impact we predict to happen via a control restoration mechanism. Below, we build on research related to social density, the psychological drivers of WOM, and compensatory control to develop our predictions. Social Density and Perceived Control In this research, we focus our attention on social density: the number of people in a given area (typically indexed as the number of people per unit area; Baumeister and Vohs 2007; Graumann and Moscovici 1986; Oldham and Rotchford 1983). As mentioned, in this research we treat density as being synonymous with social density (e.g., a high-density environment is an environment characterized by high levels of social density—i.e., by a large number of people). Research has documented a variety of psychological consequences stemming from social density, including sensory overload (Milgram 1970), hostility (Griffitt and Veitch 1971), and antisocial behavior (Zimbardo 1969). In addition to these outcomes, research in social and environmental psychology has suggested that social density may also have control-debilitating effects (Baum and Valins 1977), whereby individuals experience a perceived reduction in their personal control, manifested in their inability to perform desired actions (Baum and Valins 1977; Rodin 1976). Moreover, dense environments are often perceived as uncontrollable (Rodin 1976; Sherrod and Cohen 1978). Density affects one’s sense of personal control also because the presence of many other people may make the environment unpredictable (Cohen 1978), as well as interfere with the achievement of an individual’s goal(s) (Rodin 1976; Sherrod and Cohen 1978). Indeed, research has revealed that living in dense spaces leads individuals to experience a sense of reduced control over the environment; for instance, students living in dormitory apartments with high (vs. low) person-per-room density feel less personal control (Rodin 1976). Similarly, when larger (vs. smaller) numbers of dormitory residents share common spaces, these residents perceive their interactions with others as more unpredictable and, as a result, experience a loss of perceived control (Baum and Valins 1977). A similar negative effect of social density on perception of control has been suggested also in other contexts, such as densely populated urban environments (Cheng 2010; Jacobs and Appleyard 1987; Wirth 1938), as well as dense retail and service environments (van Rompay et al. 2008). In sum, substantial research supports our prediction that as an environment gets denser (i.e., as the number of people in a given space increases), individuals in that environment might perceive a loss of personal control. Perceived Control and WOM Building on the established finding that individuals like to feel in control over their environments (Rothbaum, Weisz, and Snyder 1982; Skinner 1996), as this feeling makes them perceive that they are the agent of their actions (Wegner and Wheatley 1999), past research has demonstrated that when individuals experience a threat to their sense of control, they are likely to engage in compensatory behaviors aimed at restoring it (Kay et al. 2010; Langer 1975; Whitson and Galinsky 2008). Indeed, personal control is arguably one of the most fundamental human needs: from an evolutionary standpoint, individuals who have control over the environment have a far better chance of survival (Mittal and Griskevicius 2014). Furthermore, a sense of personal control is related to people’s desire for stability, predictability, and safety (Staub, Tursky, and Schwartz 1971), which are ranked among the most basic—and thus urgent—human needs, even before belongingness or self-esteem (Maslow 1943). Thus, a lost sense of control can be considered a strong primary driver of individuals’ compensatory behaviors. Marketing research supports the idea that individuals tend to engage in control-restoring behaviors. For example, when consumers perceive a loss of personal freedom, they seek to reestablish it by making more varied and unique choices (Levav and Zhu 2009). Additionally, a reduction in consumers’ sense of control makes them more likely to engage in compensatory behaviors such as buying “structured” products (e.g., products with sharp edges or tangible boundaries; Cutright 2012) or preferring brand extensions that have a high degree of fit with the parent brand (Cutright et al. 2013). In this research, we propose that WOM is another important means by which consumers can compensate for their threatened sense of control. Past work seems supportive of our idea of compensatory WOM, as it suggests that WOM is often the expression of compensatory behaviors triggered by different needs (Berger 2014). For example, when their self-esteem is threatened, consumers are likely to engage in WOM as a way to self-enhance and restore a sense of self-worth (De Angelis et al. 2012). Moreover, when consumers have a negative service or product experience (e.g., a delayed flight), they reduce the resulting negative feelings (e.g., anger) by venting (i.e., sharing their experience with others; Nyer 1997). WOM can also facilitate social bonding, allowing individuals to compensate for a sense of loneliness and social exclusion via information sharing (Berger 2014). Extending these findings, we propose that WOM can additionally serve to mitigate a loss of personal control. In fact, information sharing can restore control for a number of intertwined reasons. First, consistent with its impression management function (Berger 2014), WOM is an important means of expressing one’s personality in social contexts (Belk 2013; Berger 2014; Dichter 1966; Sedikides 1993; Sirgy 1982), which allows individuals to exert control over their own image and affirm themselves. In fact, personal control is an important component of self-integrity (Sherman and Cohen 2006): when personal control is threatened, self-affirmation allows people to restore their view of being in control of their life’s outcomes (Liu and Steele 1986). Second, consistent with the social bonding function of WOM (Berger 2014), information sharing can strengthen social connections (Dunbar and Dunbar 1998), and people with tighter social connections can better exercise control over the events they face (Cohen 1981). In line with this idea, individuals with a strong social support system are less likely to seek further compensatory sources of control when their personal control is threatened (Cutright 2012). Third, consistent with the persuasion function of WOM (Berger 2014), individuals often engage in WOM to advise others regarding how to behave in certain situations (Fitzsimons and Lehmann 2004; Peluso et al. 2017), thus helping them make better buying decisions (Hennig-Thurau et al. 2004; Sundaram et al. 1998). By sharing useful information, consumers affirm a positive image of themselves as persons capable of influencing others (Barasch and Berger 2014), which may in turn increase their perception of control (Sommer and Burgeois 2010). Indeed, recent research suggests that individuals can regain a lost sense of control via advice giving, whereby advisors exert influence over—and thus feel in control of—advisees’ actions (Peluso et al. 2017). In sum, previous research provides a foundation for our hypothesis that consumers can instrumentally apply WOM to compensate for a feeling of low control induced by being in dense environments. OVERVIEW OF STUDIES To test our predictions, we conducted six studies. Studies 1 and 2 test whether social density is positively related to real information-sharing behavior in real-life settings. Study 3 explores the robustness of these findings in an experimental setting in which social density is manipulated. Study 4 shows that the chronic need for control moderates the effect of social density on information sharing, thereby offering first evidence for the underlying mechanism. Study 5 demonstrates that the effect of social density on information sharing is nulled when participants have the opportunity to restore control before they express their intentions to engage in WOM. Finally, study 6 tests whether consumers perceive that sharing information does, in fact, restore control in high-density settings. In our empirical tests, social density is observed in an urban setting (study 1), manipulated in laboratory (studies 3 and 4) or online (study 6) experiments, or rated (by an external observer in study 2, and by participants in study 5). ADDITIONAL EMPIRICAL QUESTIONS In addition to the main goals above, and further contributing to research on social density, our empirical investigation also addresses the role of physical proximity. Social density and physical proximity are often confounded, because typically as the number of individuals in a given space (i.e., social density) increases, so does the physical proximity between them. However, it is important to notice that social density and physical proximity are conceptually distinct and not necessarily correlated, as physical proximity may vary independently of variations in social density (Grauman and Moscovici 1986; Paulus et al. 1976). For example, given the same physical proximity between people, by definition social density is greater when more (vs. fewer) people are in the same space. Conversely, given the same level of social density (i.e., the same number of people in a given area), individuals can be more (vs. less) close to each other. Thus, in theory, an effect of social density on WOM could emerge independent from physical proximity (and vice versa). When social density and physical proximity are confounded, it is not possible to conclude that any observed effect on WOM is due to social density. Indeed, when social density and proximity go hand in hand, it would be entirely possible that physical proximity drives any observed effect on WOM, rather than social density per se. If this were the case, it would mean that an effect on WOM could in fact emerge when a person is physically close to another, regardless of whether the environment is dense (i.e., independent of the number of people in the immediate surroundings; for example, when just one person is standing too close), and would not emerge in a dense room if people were reasonably far apart. In other words, it would be entirely possible that social density would demonstrate no effect at all. Thus, to ensure the validity of our main thesis, we investigate whether an effect of social density emerges independent from physical proximity. We address this issue in studies 2 (in which we statistically control for physical proximity), and 3 (in which we experimentally control for it). Beyond physical proximity, we also explore the role of affect and arousal in studies 3 and 4. Indeed, research indicates that both social density and control loss are sometimes (but not always) associated with negative arousal (Averill 1973; Folkman 1984; Griffitt and Veitch 1971; Paulus et al. 1976; Paulus and Matthews 1980). Therefore, it is important to investigate whether negative arousal contributes to the effect of social density on WOM, or whether such an effect emerges even when the negative affective consequences of social density do not materialize. STUDY 1: URBAN DENSITY AND WOM In this pilot study, we explored whether social density is related to WOM in real life. Specifically, because research shows that social density effects are observable at the urban level (Milgram 1970; Steblay 1987), in this study we tested whether there is a positive relationship between objective urban density (= city population size/city area) and real online WOM. To achieve this goal, we monitored information spread via the online microblogging platform Twitter, in which users share content in the form of brief texts, called tweets. Indeed, microblogging has emerged as one of the most used online tools for WOM, and Twitter is one of the most popular information-sharing platforms (Jansen et al. 2009; Wolny and Mueller 2013). We recorded the tweets shared in a sample of Italian cities and tested whether urban density was correlated with number of tweets per capita in these cities. Procedure Our dataset included 1,421 Italian cities, each of which had more than 5,000 inhabitants, according to current census data (ISTAT). We monitored Twitter activity in these cities daily over a period of two months through FindTheRipple (an Artificial Intelligence–driven platform analyzing social media activity; findtheripple.com), and recorded the number of tweets shared in each of these cities, using city coordinates to detect geolocalized tweets. Overall, we collected 2,194,856 geolocalized tweets from 203,071 active users. We subsequently excluded data from 14 cities in which FindTheRipple detected unusual Twitter activity, due to the presence of Twitter celebrities sharing an abnormal number of tweets (studentized deleted residuals > 4; final N = 1,407). Results Tweets per Active User Urban density (= city population size/city area, according to current census data; ISTAT) and the number of tweets per active user shared daily in each city were significantly correlated (r = .11, p < .001). We also regressed city population size and city area on tweets per active user in a first step of a multiple regression procedure (Aiken and West 1991), R2 = .01; F(2, 1,404) = 5.21, p < .01. This analysis revealed a main effect of population size on tweets per active user (β = .09, t(1,404) = 2.95, p < .01); there was no main effect of city area. The interaction between population size and city area entered in a second step proved significant and improved the main effects model (β = –.11, t(1,403) = –2.28, p < .05; Fchange(1, 1,403) = 5.21, p < .05). This interaction indicates that the effect of population size on tweets per active user is stronger as the city area decreases; in other words, as urban density (i.e., the number of inhabitants per unit space) increases, so does information sharing. Of note, the effects outlined above held also after we controlled for the number of active users in each city. In conclusion, this pilot study suggests that the main driver of the observed effect is the number of people in the physical environment of a focal consumer, and that the smaller the area of this physical environment, the greater this effect becomes. In other words, social density (the number of people in a given area) is positively related to information sharing. STUDY 2: CAMPUS DENSITY AND WOM Study 1 provided initial correlational evidence that social density is related to real information sharing at the urban level. In study 2, we tested whether a similar relationship could be observed in a different naturalistic setting on a smaller scale and relatively more controlled setting. In this study, we used an external observer (a research assistant) to rate social density and we tested whether social density (as rated by an external observer) would correlate with real information sharing. We recruited students on one campus of Nova University Lisbon, collecting data on different days of the week (Monday to Saturday) and at different times of the day (early afternoon, evening, and late evening) in order to capture the varying levels of social density during peak and lull hours. In order to have a proxy of real sharing behavior, respondents read a short article and were led to believe that, by checking a box during the survey, they would automatically share this article on their favored social network. In addition to its two main purposes, this study also provides a first test of the role of physical proximity. In fact, proximity between people may or may not increase in tandem with social density; thus, it is important to know whether an effect of social density on WOM can emerge independent from people’s physical closeness. Thus, the same external observer rated the proximity between the respondent and other people, and we controlled for this variable statistically in our analysis. Procedure Three hundred eighteen respondents took part in this study on a voluntary basis. A research assistant—aware of the definition of social density but blind to the hypotheses—approached respondents on one campus of Nova University Lisbon. The research assistant and respondents used a tablet to fill in the survey. First, unbeknownst to respondents, the research assistant rated how densely populated each respondent’s immediate surroundings were (1 = Not at all, 7 = Extremely), as well as how physically close the respondent was to other people (1 = Not at all close, 7 = Very much). Next, the assistant handed the tablet to the respondents, who read a short article about Lisbon (appendix A), and were led to believe that by checking a box, they would share this article on a social network of their choice. After filling in their demographic information, respondents were informed that they would actually not share the article. Finally, they were debriefed and thanked for their participation. Results Social Media Sharing We entered social density and physical proximity (as rated by the research assistant) as predictors in a logistic regression model that predicted social media sharing (1 = Yes, 0 = No). We added the interaction between social density and physical proximity in a second step. The interaction term in the second-step model did not improve the main effects model (χ2(1, N = 303) = 0, p = .99) and thus will not be discussed further. Most importantly, the main effects model (χ2(2, N = 303) = 10.99, p < .01) revealed a significant effect of both social density and physical proximity. Interestingly, physical proximity had a negative effect on information sharing (B = –.19, χ2(1) = 6.14, p = .01)—a result that aligns with the idea that violations of physical space might lead to antisocial behaviors (Burgoon 1993; Zimbardo 1969) and thereby decrease the social transmission of information. Most important, and in line with our hypotheses, social density had a positive effect on information sharing (B = .25, χ2(1) = 6.41, p = .01), even after we controlled for the negative effect of physical proximity. In sum, this study offers converging evidence for our proposed effect of social density on WOM in a real-life context. This study also revealed that the effect of social density on WOM can emerge above and beyond that of physical proximity. Indeed, we disentangled the effect of physical proximity from that of social density, and uncovered that the latter effect is positive and significant even after the former is statistically controlled. Finally, the results of studies 1 and 2 have significant practical implications, as they revealed an effect of social density on real information sharing or on a proxy of it, respectively. Moreover, social density was either measured objectively (study 1) or rated by an external observer (study 2). This suggests the interesting possibility that companies could rely on their own measurements and target consumers with shareable content when they can infer that the immediate environment is highly dense. STUDY 3: MANIPULATED SOCIAL DENSITY AND WOM Studies 1 and 2 are important because they allowed us to observe that social density is related to information sharing in real life. The correlational nature of this evidence, however, might be problematic, because it does not allow us to rule out that unobserved variables effectively drive the reported effects. For example, it is possible that people who live in densely populated areas differ from those who live in less densely populated areas on a variety of characteristics that affect Twitter usage; or that students who are on campus during peak (vs. lull) hours are more extraverted, and thus more likely to communicate with others. Therefore, the first goal of study 3 was to rule out these potential alternative explanations, by employing an experimental procedure in which we randomly assigned participants to high- and low-density environments. Consistent with its definition, we manipulated social density by varying the number of individuals in a fixed space. Moreover, whereas in study 2 we controlled for individuals’ physical proximity in our statistical analysis, in this study we controlled for physical proximity experimentally by keeping it constant between conditions; thus, we offer more compelling evidence that the effect on WOM is attributable to social density (the number of individuals in a given space) and not to individuals’ physical closeness. In addition to controlling for physical proximity, we explored the role of emotions and arousal as potential concurrent mechanisms. In fact, individuals in dense settings sometimes experience negative affect (Griffitt and Veitch 1971) and therefore might try to restore their positive mood by seeking a connection with others (Gray, Ishii, and Ambady 2011). Alternatively, the greater arousal in a dense setting might trigger information sharing (Berger 2011). Procedure Eighty-six students (58.1% male, Mage = 20.19, SD = 1.38) participated in this study in exchange for course credit. Students were invited to participate in groups of 25 individuals at a time and assigned to one of two social density (high vs. low) conditions. In the high-density condition, all attending participants were seated in the same 24-person classroom; in the low-density condition, participants were divided into two adjacent identical 24-person classrooms (these two rooms had equal size). In order to hold physical proximity constant, participants were assigned to adjacent and equally spaced seats in both conditions. Participants in both conditions completed a paper questionnaire. First, participants completed a battery of filler questions, so that they would spend at least some time in a high-density versus low-density setting before they completed the dependent measure. Next, participants read an article about the Jawbone UP3, a fitness-tracking product (appendix B), and indicated how likely they would be to share this content with other people (1 = Very unlikely, 7 = Very likely). Subsequently, they completed a short version of the PANAS (Watson, Clark, and Tellegen 1988), which measured their current emotions, as well as an anxiety measure (Spielberger et al. 1983) and an arousal measure (Thayer 1989). Next, participants rated the extent to which the room was noisy (1 = Not at all, 7 = Extremely), as well as the proximity between themselves and others (1 = Very close, 7 = Very far). Finally, participants were debriefed and thanked for their participation. Results In this and all following analyses, we excluded missing cases pairwise. Manipulation Check Social density (indexed as the number of people per unit space; Baumeister and Vohs 2007) was higher in the high- (M = .43, SD = .11) versus low-density condition (M = .25; SD = .04; t(82) = 10.13, p < .001). Proximity between participants was similar between conditions as intended (p > .40); however, even though noise was rather low in both conditions, the high-density condition was still noisier (M = 3.44, SD = 1.66) than the low-density condition (M = 2.56, SD = 1.25; t(84) = 2.82, p < .01). There were no significant differences in negative affect, positive affect, anxiety, and arousal between conditions (p > .54). Likelihood to Share Information Participants in the high-density condition indicated that they were more likely to share the content of the article with others (M = 3.61, SD = 1.68) than were participants in the low-density condition (M = 2.80, SD = 1.70; t(84) = 2.19, p < .05). Overall, these results suggest that social density triggers information sharing. Moreover, the effect does not seem to be generated by affect, anxiety, or arousal, as these factors were similar between conditions. Moreover, because proximity was constant between conditions, this study also suggests that the effect of social density on WOM is driven by the number of people in a given space rather than by the proximity between these people. Together, these insights suggest that an effect of social density on WOM can emerge even in situations in which physical proximity and negative affective states are not involved. STUDY 4: NEED FOR CONTROL MODERATES THE EFFECT OF SOCIAL DENSITY ON WOM In study 3, we experimentally controlled for physical proximity. However, we acknowledge that real-life social density is often associated with greater physical proximity. Thus, in an effort to achieve higher ecological validity, we allowed physical proximity to correlate with social density in this experiment. Whereas in study 3, participants’ seats were equally spaced in both experimental conditions, in study 4 participants were seated closer to each other in the high- (vs. low-) density condition. Moreover, because noise was correlated with social density in study 3, we kept noise constant between conditions in this study. We achieved this goal by inviting fewer participants at a time—as compared to study 3—such that they would generate very little noise, even in the high-density condition. Most importantly, the goal of this study was to investigate the proposed mechanism. Earlier, we reasoned that the effect of social density on WOM is driven by the desire to compensate for a perceived loss of control. Therefore, the effect of social density should be greater among those individuals who have a higher (vs. lower) chronic need for control. In other words, we predicted that the effect of social density on the likelihood to share information would be moderated by participants’ trait need for control. Procedure One hundred nineteen students participated in this study in exchange for course credit (40.2% male; Mage = 20.24, SD = 1.82). We invited eight students at a time and assigned them to one of two social density (high vs. low) conditions. In the high-density condition, all attending participants were seated in a four-person lab room; in the low-density condition, participants were seated in two adjacent identical four-person lab rooms (these two rooms had equal size). Unlike study 3, in which participants’ seats were equally spaced in both conditions, participants’ seats were closer in the high-density condition than in the low-density condition. First, participants read the same article as in study 3 and reported the likelihood that they would share that article with others (1 = Very unlikely, 7 = Very likely). Next, they completed a short version of the PANAS (Watson et al. 1988), which measured their current emotions, as well as an anxiety measure (Spielberger et al. 1983) and an arousal measure (Thayer 1989). Subsequently, participants completed a three-item scale that measured their chronic need for control (e.g., “I like to be in control of most things that occur in my life”; 1 = Strongly disagree, 7 = Strongly agree; Burger and Cooper 1979; α = .76; appendix C). Then, they rated the noisiness of the room, as well as the proximity between themselves and others, as in study 3. Finally, participants were debriefed and thanked. Results Manipulation Check Social density (indexed as the number of people per unit space; Baumeister and Vohs 2007) was higher in the high-density (M = .54, SD = .05) than in the low-density condition (M = .27, SD = .04; t(116) = 33.22, p < .001). There were no significant differences between conditions in terms of noise, arousal, anxiety, and positive or negative affect (p > .35). While in study 3 the physical distance between participants was similar between conditions, participants in study 4’s high-density condition sat closer to each other (M = 5.81, SD = 1.23) than did participants in the low-density condition (M = 3.83, SD = 1.10; t(117) = 9.25, p < .001), as intended. Likelihood to Share Information We submitted this measure to a multiple regression procedure (Aiken and West 1991), in which we entered the main effects of social density and chronic need for control in a first step, and the two-way interaction between social density and chronic need for control in a second step. The first model was not significant (p > .27); however, performing the second step significantly improved the model (Fchange(1, 115) = 6.56, p = .01), as a two-way interaction between social density and chronic need for control emerged (B = 1.09, t(115) = 2.56, p = .01). We then decomposed this two-way interaction at high (+1SD) and low (–1SD) levels of chronic need for control. Participants with a high chronic need for control were more likely to share the product information when in a high-density room, compared to their counterparts in a low-density room (B = 1.27, t(115) = 2.85, p < .01). Meanwhile, there was no effect of social density on information sharing among participants with a low chronic need for control (p > .29; see figure 1). FIGURE 1 View largeDownload slide THE POSITIVE EFFECT OF SOCIAL DENSITY ON INFORMATION SHARING IS SIGNIFICANT ONLY AMONG PEOPLE WHO HAVE A HIGH CHRONIC NEED FOR CONTROL FIGURE 1 View largeDownload slide THE POSITIVE EFFECT OF SOCIAL DENSITY ON INFORMATION SHARING IS SIGNIFICANT ONLY AMONG PEOPLE WHO HAVE A HIGH CHRONIC NEED FOR CONTROL We recognize that, unlike in study 3, the manipulation of social density did not produce a main effect in this study. We present an additional analysis of this study’s data and further speculate on potential boundaries conditions for a main effect of social density in the General Discussion. Most importantly, study 4’s results are consistent with the proposed process—namely, that the chronic need for control moderated the effect of social density on WOM. Finally, this study allowed us to rule out some alternative mechanisms, such as the effect of environmental noise, which we controlled for by keeping noise constant between conditions. Moreover, as in study 3, negative affect, positive affect, anxiety, and arousal were similar between conditions; thus, they are unlikely explanations for the observed effects. Also as in study 3, the degree of interpersonal interaction was constant between conditions, as participants were not allowed to interact during the study; therefore, a lack of interpersonal interaction is an unlikely explanation for the observed differences between the high- (vs. low-) density conditions. STUDY 5: ALTERNATIVE MEANS TO RESTORE CONTROL REDUCE THE EFFECT OF SOCIAL DENSITY ON WOM If people in high-density settings engage in WOM as a way to compensate for the loss of control they experience, then providing an alternative means of restoring control beforehand should negate the positive effect of social density on WOM. We reasoned that once a source of control has been provided, the effect of social density on subsequent information sharing should be reduced or disappear, because the need to restore control should have already been replenished. In fact, research shows that sources of control can substitute for each other (Inesi et al. 2011; Kay et al. 2009). Thus, study 5 tests whether the likelihood of engaging in WOM is attenuated when people have an alternative way of restoring their threatened sense of control. We used a false feedback paradigm to manipulate the opportunity to restore control. We gave participants an easy interactive puzzle, in which they had to click on a series of dots numbered from 1 to 66 to reveal a hidden picture. We told participants they would see a clearer picture solution based on their accuracy in clicking all 66 dots. In reality, participants were assigned to one of two conditions, regardless of their real performance. In the control over outcome condition, participants were told that they were accurate, and could see a picture of a dog as the solution to the puzzle; in the no control over outcome condition, participants were told that they were not accurate, and could not see the correct solution to the puzzle (they saw an unintelligible scribble instead; appendix D). Assuming that participants put the same effort into the task between conditions, we expected that participants who saw the scribble outcome would perceive less control over the outcome than participants who saw the correct solution to the puzzle. In addition to manipulating the opportunity to restore control, study 5 had three goals. First, we wanted to test the effect of social density in yet another naturalistic setting. Thus, we recruited participants from an online subject pool (Amazon Mechanical Turk, henceforth MTurk) and required them to be in a public space and use a GPS-enabled hand-held device (such as a smartphone or a tablet) so that we could verify their approximate location. Next, participants were instructed to focus on the environmental features of their current location (its area and the number of people in that area) and subsequently rate its social density on a one-item scale—our focal measure of social density, validated in previous research (Kalb and Keating 1981; see also the web appendix, validation study 1). Second, we used a different measure of information sharing, as well as different shareable content, in order to test the robustness of our results. Third, we wanted to rule out a possible explanation for the effect; specifically, one might argue that individuals who are in a high-density location focus their attention on the content on their mobile devices in order to distract themselves and find relief from their highly dense surroundings (Andrews et al. 2015). Thus, individuals in high-density locations might share content simply because they have paid greater attention to it. Therefore, we measured how much time participants devoted to reading an article as an indication of the extent to which they attended to the information in said article. Procedure One hundred sixty-four US residents (52.4% male; Mage = 30.92, SD = 20.01) from an online pool (MTurk) took part in this study in exchange for monetary compensation. Participants were required to complete this survey in a public space, using a GPS-enabled hand-held device. Participants were informed that their location and type of device would be detected. Participants’ location latitude and longitude, as well as their device type, were recorded via metadata features of the online data collection software (Qualtrics). First, participants reported their current location (bar, library, café, restaurant, etc.) and the address of that location. Next, participants focused on the environmental features of their current location by estimating the number of people in their surroundings and the approximate size of their location, and subsequently rated how densely populated their location was (1 = Not at all, 7 = Extremely; our focal measure of social density, based on Kalb and Keating 1981). Of note, participants’ ratings of social density are our focal independent variable—in lieu of participants’ estimates—because these previously validated ratings (Kalb and Keating 1981) more closely correlate with objective social density (see the web appendix, validation study 1). After this battery of questions, participants completed a connect-the-dots puzzle, consisting of a collection of 66 dots numbered from 1 to 66. The goal of this puzzle was for participants to reveal a hidden picture by clicking the mouse or tapping a finger on each dot, starting with dot 1 and ending with dot 66. Participants were told that after they completed the task, our software would draw their solution, taking into account all of their moves and overall accuracy. In reality, after completing this task, participants were assigned to one of two conditions, regardless of their performance: in the control over outcome condition, participants were told that they were accurate, and they could see that they drew a picture of a dog; in the no control over outcome condition, participants were told that they were not accurate and were shown an unintelligible scribble as their drawing (appendix D). Unbeknownst to participants, we registered both the number of clicks and the amount of time that participants spent on the task, in order to verify the important assumption that participants put equal effort in the task between conditions. Following the control manipulation, participants read an article about an innovative tap design (appendix E). We unobtrusively measured the time that participants spent reading the article. Next, participants answered two questions that measured their likelihood of talking about the article with others (“How likely are you to share this content with others on social media?” and “How likely are you to talk about this information with others?” 1 = Very unlikely, 7 = Very likely; r = .71, p < .001). Participants also indicated which action they would be more likely to do: share this article on their Facebook timeline, share this article in a private message on Facebook, click “Like” but not share this article, or none of these options. Next, participants indicated whether they were Facebook users and completed a manipulation check (“In the dot-to-dot puzzle, how much control did you have over the final outcome?” 1 = No control at all, 7 = Full control). Finally, participants were asked to recall the feedback they received on their performance, then debriefed and thanked for their participation. Results The following analyses include only participants who completed the manipulation task with sufficient accuracy. Because the manipulation task required participants to click on 66 dots, we excluded 25 participants with fewer than 60 clicks. The percentage of excluded participants (15.6% in the control over outcome condition, 14.9% in the no control over outcome condition) was not significantly different between conditions (χ2(1, N = 164) = .01, p = .91). Manipulation Check Participants in the control over outcome (M = 157.87, SD = 88.12) and the no control over outcome condition (M = 157.47, SD = 67.96) spent an equal number of seconds on the task (t(137) = –.03, p = .98). Moreover, participants in the former condition (M = 68.95, SD = 6.63) clicked a similar number of dots as those in the latter condition (M = 69.34, SD = 9.21; t(137) = .29, p = .77). Thus, it is reasonable to assume that participants put a comparable amount of effort into the task between conditions. Most importantly, participants in the control over outcome condition perceived that they had more control over the task outcome (M = 5.86, SD = 1.52) than did participants in the no control over outcome condition (M = 3.65, SD = 1.61; t(137) = 8.37, p < .001), as intended. All but four participants could recall the outcome of the task and feedback they received (excluding these participants does not change the patterns and significance of results). A validation study (see the web appendix, validation study 2) confirmed that the manipulation affected perceived control (the same pattern and significance of results as above emerged for the manipulation checks), but not self-efficacy (Bandura 1994) or self-esteem (Rosenberg 1989). Reading Time Participants in the control over outcome condition (M = 70.91, SD = 132.65) spent approximately the same amount of time reading the article as participants in the no control over outcome condition (M = 64.58, SD = 113.32; t(137) = .30, p = .76). This suggests that the level of attention and the depth of elaboration were similar between conditions. Moreover, social density was not related to reading time, nor was there an interaction effect between social density and the control manipulation on reading time (p > .26). Likelihood to Share Information We averaged the two sharing likelihood items to form a composite measure of likelihood to share information. We applied a multiple regression procedure to analyze this composite measure (Aiken and West 1991). The main effects model was not significant (R2 = .01; F(2, 136) = .80, p = .45); however, adding the interaction between social density and the control condition significantly improved the model (Fchange(1, 135) = 5.82, p < .05), as a significant interaction emerged (B = –.41, t(135) = –2.41, p < .05). We followed up this significant interaction by analyzing the effect of social density in both conditions. As predicted, there was a positive effect of social density on sharing when participants did not have an alternative means of restoring control (B = .32, t(135) = 2.62, p < .01). When an alternative means was available, the effect of social density on sharing was not significant (B = –.08, t(135) = –.73, p > .46; see figure 2). Our analysis of the single items revealed a similar pattern of results. FIGURE 2 View largeDownload slide THE POSITIVE EFFECT OF SOCIAL DENSITY ON INFORMATION SHARING OCCURS ONLY WHEN PARTICIPANTS DO NOT HAVE AN ALTERNATIVE MEANS TO RESTORE CONTROL FIGURE 2 View largeDownload slide THE POSITIVE EFFECT OF SOCIAL DENSITY ON INFORMATION SHARING OCCURS ONLY WHEN PARTICIPANTS DO NOT HAVE AN ALTERNATIVE MEANS TO RESTORE CONTROL Facebook Sharing We excluded 22 people who did not use Facebook from this analysis. The percentage of excluded participants (16.1% in the control over outcome condition, 10.4% in the no control over outcome condition) was not significantly different between conditions (χ2(1, N = 164) = 1.14, p = .29). Based on a validation study (see the web appendix, validation study 3) showing that people perceive the act of posting on a Facebook timeline or sending a message as means to share information with others, but do not perceive “liking” as such, we created a new variable—Facebook sharing—by coding the first two choice options (sharing on Facebook timeline and sharing in a private Facebook message) as 1 (= Sharing), and the last two options (liking but not sharing, and doing nothing) as 0 (= Not sharing). We ran multiple logistic regressions to analyze this variable. This analysis revealed that adding the interaction between social density and the opportunity to restore control improved the main effects model (χ2(1, N = 119) = 2.82, p = .09). Indeed, there was a marginally significant interaction between the two variables (B = –.39, χ2(1) = 2.78, p = .10). Decomposing this interaction revealed that social density had a positive effect on sharing probability when participants did not have an alternative means of restoring control (B = .34, χ2(1) = 3.80, p = .05), but this effect became nonsignificant when an alternative means of restoring control was available (B = –.05, χ2(1) = .10, p = .75; see figure 3). FIGURE 3 View largeDownload slide SOCIAL DENSITY BOOSTS THE PROBABILITY TO SHARE CONTENT ON FACEBOOK ONLY WHEN PARTICIPANTS DO NOT HAVE ALTERNATIVE MEANS TO RESTORE CONTROL FIGURE 3 View largeDownload slide SOCIAL DENSITY BOOSTS THE PROBABILITY TO SHARE CONTENT ON FACEBOOK ONLY WHEN PARTICIPANTS DO NOT HAVE ALTERNATIVE MEANS TO RESTORE CONTROL In sum, study 5 revealed that the effect of social density in WOM was attenuated when consumers had (vs. did not have) an alternative opportunity to restore control. We have interpreted this result as converging evidence that the need to restore control underlies the effect of social density. STUDY 6: WOM RESTORES CONTROL IN HIGHLY DENSE ENVIRONMENTS In study 6, we tested the core of our conceptual model: Do high-density environments in fact instill a perceived loss of control? And does engaging in WOM in high-density settings help consumers restore their sense of control? In this study, we tested whether consumers perceive that sharing information mitigates the loss of perceived control that they experience in high- (vs. low-) density environments. Specifically, we presented participants with vignettes in which they shared (vs. did not share) an article on social media while they were either in a more (vs. less) densely populated subway train car, and we measured their perceived control in that situation. Procedure Four hundred thirty-three US residents from an online pool (MTurk) participated in a 2 (social density: high vs. low) × 2 (WOM: sharing vs. not sharing) between-participants experiment (39% male, Mage = 37.24, SD = 22.53) in exchange for monetary compensation. First, participants in the high- [low-] density condition read the following: “You are sitting in a 200-person subway train car during the day and there are about 120  people around you. You are reading an interesting article on your smartphone.” Next, participants were assigned to a WOM condition. Participants in the not sharing condition indicated how much control they would feel in the aforementioned situation by completing a four-item perceived control measure (ad hoc; e.g., “I would have control over my surroundings”; 1 = Strongly disagree, 7 = Strongly agree; appendix F) immediately after the scenario. Participants in the sharing condition, by contrast, read the following statement before completing the same four-item perceived control measure: “You decide to share this article with your contacts on Facebook. You click ‘share’ and it’s done! You have shared the article on your Facebook page for your contacts to read.” Because the success of our manipulations hinged upon participants reading the scenarios and noticing the social density and WOM details, all participants completed an attention check in which they indicated how many people were in their subway car in the text they had just read (three-point scale: About 10/About 120/I do not remember or I am not sure), and whether they shared an article on Facebook in that scenario (three-point scale: Yes/No/I do not remember or I am not sure). Finally, participants indicated whether they were Facebook users, filled in their demographic information, and were debriefed. Results As in study 5, this analysis excludes 28 participants who were not Facebook users. It also excludes 82 participants who failed the attention check. The percentage of excluded participants was similar between conditions (< 27.8%; χ2(1, N = 433) = .39, p = .53). Perceived Control We submitted the perceived control data to a two-way ANOVA in which social density and WOM served as factors. There was a significant main effect of social density condition on perceived control (F(1, 320) = 13.27, p < .001): overall, participants who imagined a high-density subway car indicated that they would feel less control (M = 4.57; SD = 1.38) than participants who imagined a low-density subway car (M = 5.16; SD = 1.10), as hypothesized. Also consistent with our theorizing, there was a significant main effect of WOM (F(1, 320) = 19.08, p < .001): overall, participants who imagined that they engaged in WOM indicated that they would perceive more control (M = 5.11, SD = 1.10) than participants who did not imagine engaging in WOM (M = 4.60, SD = 1.41). These main effects were qualified by an interaction between social density and WOM (F(1, 320) = 3.99, p < .05). Specifically, a simple main effects analysis revealed that engaging in WOM (vs. not) did not increase perceived control among participants who imagined themselves in a low-density car (p = .25), since they already perceived themselves as being in control. However, participants in the high-density condition who imagined engaging in WOM reported higher levels of perceived control (M = 4.95, SD = 1.15) than their counterparts who did not imagine engaging in WOM (M = 4.18, SD = 1.49; p < .001; see figure 4). Thus, consumers perceive that high-density environments reduce their control, and that engaging in WOM—in this case, by sharing information on Facebook—restores their lost sense of control. FIGURE 4 View largeDownload slide SHARING INFORMATION HELPS CONSUMERS IN HIGH-DENSITY LOCATIONS TO RESTORE THEIR LOST SENSE OF CONTROL FIGURE 4 View largeDownload slide SHARING INFORMATION HELPS CONSUMERS IN HIGH-DENSITY LOCATIONS TO RESTORE THEIR LOST SENSE OF CONTROL GENERAL DISCUSSION In this research, we tested the prediction that consumers in high-density settings engage in WOM as a means of compensating for their perceived loss of control. To test our hypotheses, we conducted one pilot field study with an objective urban density measure and real online sharing behavior (study 1), one survey-based field study (study 2), two laboratory experiments (studies 3 and 4), and two online experiments (studies 5 and 6). Study 1 and 2 found support for the basic effect in two different real-life settings. In study 1, we observed a correlation between urban density and real information sharing (number of shared tweets per capita) in a sample of Italian cities. This study revealed that urban density is positively correlated with information sharing. Extending these results to another context, study 2 revealed a correlation between social density (rated by an external observer) and a proxy of real sharing behavior among students on a Portuguese university campus. Study 3 found support for the basic effect in an experimental setting: participants placed in a high- (vs. low-) density room expressed a higher likelihood of sharing information with others. Moreover, study 3 empirically disentangled the effect of physical proximity between people from that of social density, demonstrating that the latter has a positive effect on WOM even when we experimentally control for the former. Thus, this result complements study 2—in which we statistically controlled for physical proximity—and suggests that physical proximity is not the main driver of the observed effects. In sum, the current research suggests that the crucial factor driving the effect of social density on WOM is the number of people in a focal consumer’s immediate surroundings, rather than the physical proximity between this focal consumer and other people. In study 4, we provided initial evidence for the underlying process by linking our findings with chronic need for control. We found that social density increased the likelihood of information sharing among participants who scored high on a chronic need for control scale, but it did not exert this effect among participants with low chronic need for control. In study 5, we manipulated the opportunity to restore control before participants expressed their intention to share information with others. When participants did not have an alternative means of restoring control, we replicated our prior results; however, when participants could restore control via alternative means, we found that social density no longer had an effect on participants’ likelihood of sharing information because they had already satisfied their need to restore control. Meanwhile, study 6 provided evidence that social density decreases their control, and that sharing information in highly dense environments reestablishes their shaken sense of control. In other words, this study suggests that consumers perceive that WOM sharing works as a compensatory behavior in high-density settings. Theoretical and Managerial Implications From a theoretical standpoint, this research offers some relevant contributions to the WOM and compensatory control literatures. First, it sheds light on how physical context affects the likelihood of sharing information. Second, we found that social density is a factor that induces consumers to restore control, and that a previously overlooked motivation (the need to restore control) underlies WOM. Finally, while past work in marketing has explored the effect of social density on purchase-related variables (Andrews et al. 2015), our study is the first to investigate and provide evidence for the effect of social density on WOM behavior. Our work also features significant managerial implications. In the era of mobile technology and devices, companies can share content with consumers in different locations (cafés, stores, shopping malls, or other—more or less—densely populated places). Our research suggests that marketing practitioners could leverage geolocation technologies by sending targeted, real-time communications to consumers when they are in highly dense spaces, and thereby increase the effectiveness of their communication campaigns aimed at generating online or offline WOM. Moreover, in line with prior research (Inesi et al. 2011), our findings suggest that when the need to restore control is satisfied, further opportunities to restore control have diminishing returns. Thus, it is essential to time marketing actions accurately when targeting consumers in a high-density setting, as the first competitor to do so is more likely to reap the benefits. Potential Limitations and Alternative Accounts In study 5, we measured social density before our dependent variable, which might have generated some demand effect or suspicion. However, when participants were debriefed, none of them could guess our hypothesis. In the same study, participants could have been suspicious also about the false feedback they received about their performance on the connect-the-dots task, but when we debriefed participants we did not find evidence that this was the case. Thus, on the whole, demand effects are unlikely. Finally, in study 6, even though no participant could accurately guess our hypothesis, nine participants suspected that we intended to study the effect of sharing information on perceptions of control in high- versus low-density environments. However, excluding these suspicious participants did not change study 6’s results. Thus, this study’s results also are unlikely to be driven by demand effects. Alternative accounts similarly lack explanatory power in terms of our overall pattern of results. For instance, prior research suggests that cues in the environment affect the amount of WOM that specific products receive, because cues make these products accessible in consumers’ minds (Berger and Schwartz 2011). It could be argued that the mere presence of others might also act as an environmental cue that stimulates WOM. However, environmental cues have been shown to affect the amount of WOM about products specifically related to these environmental cues (i.e., what is shared), rather than the likelihood that people will share information in general. In other words, products and topics that are more frequently cued by the environment are talked about more—for example, a group of students is more likely to discuss school break travels when the room contains semester abroad books (vs. not; Berger and Schwartz 2011). If the mere presence of others in high-density environments acted as a cue in the same sense, consumers should tend to share more content that is social in nature or that contains references to a group of people (e.g., concerts). However, the products and information in our stimuli did not have a social element, so this explanation seems unlikely. Another alternative explanation is that an increasing number of people in a consumer’s immediate surroundings increases the salience of the social environment, and thus the desire to communicate. Such an effect (in isolation) would not be compatible with our empirical evidence. For example, it would be unclear why the salience of the social environment per se would lead to a higher likelihood to share information especially among individuals who have a high chronic need for control (study 4); it would also be unclear why providing an alternative means to restore control would null this effect (study 5). Instead, our theorized control-restoration mechanism accounts for results across studies more harmoniously. Moreover, past work has shown that social density might reduce social interaction in face-to-face encounters (Griffitt and Veitch 1971; Zimbardo 1969). Thus, one might argue that people in highly dense places feel the urge to share content with others (e.g., online) because social density inhibits face-to-face communication, thereby inducing both a perceived loss of control and a desire to compensate for this lack of interpersonal interaction. However, in our two controlled lab experiments (studies 3 and 4), participants did not interact with each other in either the high- vs. low-density condition; thus, the amount of face-to-face communication was similar in both studies. Thus, the inhibition of face-to-face communication in high-density conditions is an unlikely driver of the observed effects. However, we acknowledge that we could not control for nonverbal communication (e.g., eye contact, body language); this aspect is worthy of further investigation. Lastly, we recognize that, unlike in study 3, our manipulation of social density in study 4 did not produce a main effect on WOM. We conducted additional analyses to explore potential reasons for this lack of main effect. A 2 (social density: high vs. low) × 2 (study: 3 vs. 4) ANOVA with objective social density as the dependent variable revealed that social density was higher overall in study 4 (M = .39, SD = .14) as compared to study 3 (M = .34; SD = .12; F(1, 198) = 49.09; p < .001), and that the high-density condition in study 4, in particular, was significantly denser than the corresponding high-density condition in study 3 (F(1, 198) = 69.46; p < .001). Overall, these results suggest that the main effect of social density on WOM might emerge within certain social density levels (e.g., not at extreme levels of social density); therefore, it is important that further research explore these boundaries. In the section that follows, we further speculate on boundary conditions and moderators of the effect of social density on WOM. Related Research and Further Directions Prior research on compensatory control has suggested that religion (Kay et al. 2010), structure seeking (Whitson and Galinsky 2008; Friesen et al. 2014), and system justification (Kay et al. 2008) can serve as means of compensating for one’s threatened sense of control. The current research suggests that WOM can also be an important means through which individuals cope with threats to their personal control. However, while we have shown that a density-induced loss of control spurs information sharing, we have not examined which specific aspect of information sharing allows consumers to restore control. This calls for further investigation. We have suggested three potential reasons: self-affirmation, social bonding, and social influence (e.g., advice giving). Whereas recent research has provided first evidence that advice giving might serve as a means of restoring control (Peluso et al. 2017), further research should examine whether each of these WOM functions contributes to control restoration. In terms of focus, our research is specifically related to a growing literature about the effect of social density on consumer behavior. Most notably, research on social density in subway trains has shown that when passengers are in high-density cars, they are more likely to react to push messages from their mobile carriers, as compared to their counterparts in low-density cars (Andrews et al. 2015). This behavior has been attributed to “mobile immersion”: in order to escape the aversive, highly dense environment, passengers immerse themselves in mobile usage. Our research complements these findings by demonstrating another motivation triggered by social density—the need to restore control, which leads to a different behavior used to restore control, namely WOM. As a matter of fact, we kept medium usage constant in studies 2–5: all participants and respondents completed our studies using a similar medium, regardless of the level of social density. Additionally, participants interacted with the medium in a similar way, and devoted a similar amount of time to each study, across different social density levels. Thus, by design, social density did not influence involvement with mobile devices or content. In fact, study 3’s results suggest that participants paid the same amount of attention to a given text regardless of social density, which may indicate that social density does not affect people’s level of involvement with the text, or the corresponding depth of elaboration. Thus, given the same level of mobile immersion, we were able to uncover a different behavior triggered by social density. Indeed, our proposed process might have contributed to the effect observed by Andrews et al. (2015). Research suggests that consumer choice is a means through which individuals satisfy their thirst for personal control (Inesi et al. 2011). Thus, passengers in high-density trains (vs. low-density trains) might have reacted more often to mobile advertising not only because they were more involved with their mobile devices, but also because making a purchase helped them satisfy their need to reestablish control. Further research could investigate whether a loss of perceived control induced by social density also leads to a greater likelihood of making a purchase. On a different point, prior research has demonstrated that social density does not necessarily instill negative arousal. In fact, research on the effect of social density on mood indicates that negative arousal may (Griffitt and Veitch 1971) or may not arise (Paulus et al. 1976; Paulus and Matthews 1980) as a result of social density. Prior research has also revealed that a lack of personal control is not necessarily associated with stress (Averill 1973; Folkman 1984). Consistent with this prior research, we found no effect of social density on affect and arousal in studies 3 and 4; thus, overall, negative affect did not emerge as a likely underlying mechanism. Although prior research supports the idea that social density and a lack of personal control do not necessarily induce negative arousal (Averill 1973; Folkman 1984), we acknowledge that, at extreme levels of social density, a greater loss of control might indeed correlate with negative arousal. This density-induced negative arousal might attenuate (or even reverse) the effect. In fact, when social density becomes threatening, self-protection instincts might prevail and make individuals more constrained and conservative in their social behavior (Maeng, Tanner, and Soman 2013). Thus, further research should examine the potential critical levels of social density at which the effect on WOM occurs, or turns from positive to negative. We further note that in this research we have focused on social density, defined as the number of people in a given area, and—consistent with this conceptualization—we have manipulated social density by varying the number of people in a room. While we have specifically focused on the number of people in a fixed space, we recognize that there are many other ways in which space and individuals interact (Paulus et al. 1976); for instance, a consumer can share a bigger versus a smaller space with a fixed number of other people, or be located at the center versus the periphery of a group. Further experimental research could explore these other contexts. Relatedly, since study 2 unveiled a significant effect of physical proximity on WOM alongside an effect of social density, future research should also further investigate the effect of proximity on WOM, independent of social density. Furthermore, we cannot conclude whether social density needs to be perceived consciously to produce an effect on WOM (i.e., whether a similar effect could occur without conscious awareness). Similarly, it is unclear whether a loss of control in highly dense environments needs to be conscious in order to spur WOM. We can speculate that a loss of control in high-density environments can drive WOM with and without consumers’ awareness, as prior research suggests that both a conscious and an unconscious loss of control drive compensatory behaviors (Kay et al. 2008). Further research could study whether one’s level of awareness affects the link between social density, perceived control, and WOM. Moreover, while social density can be aversive, people also seek out and enjoy high-density situations such as music events (Novelli et al. 2013). These contexts suggest that whether a person identifies with people in a highly dense setting or not, or whether a person is alone or with friends, may affect the degree of control the person experiences. Further research might investigate these additional moderators for the effect of social density on WOM. Finally, while we focused on the likelihood to engage in WOM, further research could explore different outcomes—for example, how social density impacts the type and valence of shared content. From a managerial point of view, it might also be interesting to test if the medium through which information sharing occurs moderates the effect of social density on WOM; for instance, one could examine whether consumers in high- (vs. low-) density locations prefer to share information online versus face-to-face, and which of these channels is a more effective medium for consumers to restore their lost sense of control. Additionally, researchers could explore the effect of social density on the choice of online platforms to share WOM and on preferences for public versus private channels. To illustrate, people in high- (vs. low-) density settings might choose sharing platforms where they can select specific recipients of their messages (e.g., email or chats) or platforms where they can write public posts visible to all contacts (e.g., social network newsfeeds), depending on the extent to which each of these platforms allows sharers to regain control. In sum, we have provided initial evidence suggesting that social density—defined as the number of people in a given area—positively affects consumers’ tendency to engage in WOM because doing so allows consumers to restore a lost sense of personal control. These findings suggest that managers could leverage highly dense environments to increase the effectiveness of campaigns aimed at disseminating online or offline WOM (e.g., via geolocation technologies and targeted communications on mobile devices). Our findings also spur further questions, leading to interesting new avenues for research of great managerial import (what type of content is more likely to be shared, on which platforms, etc.); thus, we invite scholars to further investigate the effect of social density on information sharing. DATA COLLECTION INFORMATION The data for study 1 were collected by researchers at FindTheRipple under the supervision of the first and second authors (May through July 2015). The first author supervised the collection of data for study 2 by a research assistant at Nova University Lisbon (February through March 2017), and for studies 3 and 4 by a research assistant at RSM Erasmus University (January through April 2015). The first author collected data on MTurk for study 5 (March 2015) and study 6 (January 2018). The first author supervised the data collection for validation study 1 (web appendix), which was performed by a research assistant at Nova University Lisbon the second author collected the data for this study’s follow-up at LUISS Guido Carli University (October 2017). The first author collected data on MTurk for validation studies 2 and 3 (web appendix; January 2017 and April 2016, respectively). Study 1 data were analyzed by the first and second authors in tandem with researchers at FindTheRipple. The first and second authors analyzed the data for all other studies. The authors thank Stijn M. J. van Osselaer, researchers of the Rotterdam School of Management, Erasmus University Lunch Club, the editor, the associate editor, and three anonymous reviewers for their insightful feedback. The authors also thank Annette Bartels and Christiaan J. C. Tieman for their assistance during data collection. APPENDIX A Respondents in study 2 read the following article (adapted from Minube 2012): Lisbon is an amazing city and it enjoys more sunshine than any other European capital. It is a truly unique destination: it’s one of the oldest cities in Western Europe (more so than even Rome), the westernmost city in Europe, and home to one of the continent’s most important ports. Given its long history, it’s no surprise that Lisbon tourism offers such a wealth of things to see, from Gothic cathedrals to Postmodern art galleries, all surrounded by the Atlantic breeze and a warming sun. Need any more reasons to visit Lisbon? APPENDIX B Participants in studies 3 and 4 read the following article about Jawbone UP3, a new product that helps people keep fit (Frizell 2014): Depending on your view, Jawbone is on the road to making us all super fit athletes or brutally efficient cyborgs who operate by data alone. How many minutes of REM sleep did I get last night? What’s the difference between my heart rate while resting and during a workout? How hydrated am I? How many calories did I burn on my last run? Jawbone UP3 answers all those questions better than any other Jawbone did before. APPENDIX C Need for control scale (based on Burger and Cooper 1979): I enjoy having control over my own destiny; I like to be in control of most things that occur in my life; I prefer a job where I have a lot of control over what I do and when I do it. APPENDIX D FIGURE 5 View largeDownload slide PUZZLE SOLUTION IN THE CONTROL (VS. NO CONTROL) OVER OUTCOME CONDITION IN STUDY 5. NOTE.—ORIGINAL PUZZLE KIND COURTESY OF LA SETTIMANA ENIGMISTICA – ITALY FIGURE 5 View largeDownload slide PUZZLE SOLUTION IN THE CONTROL (VS. NO CONTROL) OVER OUTCOME CONDITION IN STUDY 5. NOTE.—ORIGINAL PUZZLE KIND COURTESY OF LA SETTIMANA ENIGMISTICA – ITALY APPENDIX E Participants in study 5 read the following article, adapted from Qiu (2015): At first glance, designing a tap that sends out jets of water in different pretty patterns may sound gimmicky and completely pointless, but there is method behind this madness. Design student Simin Qiu recognized the importance of conserving precious water and thus came up with a tap that not only is both water- and energy-saving, but also turns a stream of water into art. The design, which won an IF design concept award last year, features a double turbine that rotates as the water passes through it, creating a pretty funky lattice effect. According to Qiu’s page on Behance, this addition allows 15% of water to be saved compared with a traditional faucet. Furthermore, the temperature has been set in advance to avoid unnecessary use of electric heater switches, which helps to save energy. Users can also flick between three settings that all give different pretty patterns of water. APPENDIX F Perceived control scale (1 = Strongly disagree, 7 = Strongly agree): I would feel in control; I would have control over my surroundings; The current situation would feel out of my control; I would feel a loss of personal control. References Aiken Leona S., West Stephen G. ( 1991), Multiple Regression: Testing and Interpreting Interactions , Newbury Park: Sage. Andrews Michelle, Luo Xueming, Fang Zheng, Ghose Anindya ( 2015), “Mobile Ad Effectiveness: Hyper-Contextual Targeting with Crowdedness,” Marketing Science , 35 ( 2), 218– 33. Google Scholar CrossRef Search ADS Averill James R. ( 1973), “Personal Control over Aversive Stimuli and Its Relationship to Stress,” Psychological Bulletin , 80 ( 4), 286– 303. Google Scholar CrossRef Search ADS Bandura Albert ( 1994), “Self-Efficacy,” in Encyclopedia of Human Behavior , Vol. 4, ed. Ramachandran Vilayanur S., New York: Academic Press, 71– 81. Barasch Alixandra, Berger Jonah ( 2014), “Broadcasting versus Narrowcasting: How Audience Size Impacts WOM Valence,” Journal of Marketing Research , 51 ( 3), 286– 99. Google Scholar CrossRef Search ADS Baum Andrew, Valins Stuart ( 1977), Architecture and Social Behavior: Psychological Studies of Social Density , Hillsdale, NJ: Erlbaum. Baumeister Roy F., Vohs Kathleen D. ( 2007), Encyclopedia of Social Psychology , New York: Sage. Google Scholar CrossRef Search ADS Belk Russell W. ( 2013), “Extended Self in a Digital World,” Journal of Consumer Research , 40 ( 3), 477– 500. Google Scholar CrossRef Search ADS Berger Jonah ( 2011), “Arousal Increases Social Transmission of Information,” Psychological Science , 22 ( 7), 891– 3. Google Scholar CrossRef Search ADS Berger Jonah ( 2014), “Word-of-Mouth and Interpersonal Communication: A Review and Directions for Future Research,” Journal of Consumer Psychology , 24 ( 4), 586– 607. Google Scholar CrossRef Search ADS Berger Jonah, Milkman Katherine ( 2012), “What Makes Online Content Viral?” Journal of Marketing , 49 ( 2), 192– 205. Google Scholar CrossRef Search ADS Berger Jonah, Schwartz Eric ( 2011), “What Drives Immediate and Ongoing Word of Mouth?” Journal of Marketing , 48 ( 5), 869– 80. Google Scholar CrossRef Search ADS Bughin Jacques, Doogan Jonathan, Vetvik Ole Jørgen ( 2010), “A New Way to Measure Word-of-Mouth Marketing,” McKinsey Quarterly (April). Burger Jerry M., Cooper Harris M. ( 1979), “The Desirability of Control,” Motivation and Emotion , 3 ( 4), 381– 93. Google Scholar CrossRef Search ADS Burgoon Judee K. ( 1993), “Interpersonal Expectations, Expectancy Violations, and Emotional Communication,” Journal of Language and Social Psychology , 12 ( 1–2), 30– 48. Google Scholar CrossRef Search ADS Cheng Vicky ( 2010), “Understanding Density and High Density,” in Designing High-Density Cities for Social and Environmental Sustainability , ed. Ng Edward, London: Earthscan, 1– 17. Cohen Harry ( 1981), Connections: Understanding Social Relationships , Ames, IA: Iowa State Press. Cohen Sheldon ( 1978), “Environmental Load and the Allocation of Attention,” in Advances in Environmental Psychology , ed. Baum Andrew, Singer Jerome E., Valins Stuart, Hillsdale, NJ: Erlbaum. Cutright Keisha M. ( 2012), “The Beauty of Boundaries: When and Why We Seek Structure in Consumption,” Journal of Consumer Research , 38 ( 5), 775– 90. Google Scholar CrossRef Search ADS Cutright Keisha M., Bettman James R., Fitzsimons Gavan J. ( 2013), “Putting Brands in Their Place: How a Lack of Control Keeps Brands Contained,” Journal of Marketing Research , 50 ( 3), 365– 77. Google Scholar CrossRef Search ADS De Angelis Matteo, Bonezzi Andrea, Peluso Alessandro M., Rucker Derek D., Costabile Michele ( 2012), “On Braggarts and Gossips: A Self-Enhancement Account of Word-of-Mouth Generation and Transmission,” Journal of Marketing Research , 49 ( 4), 551– 63. Google Scholar CrossRef Search ADS Dichter Ernest ( 1966), “How Word-of-Mouth Advertising Works,” Harvard Business Review , 44 ( 6), 147– 66. Robin Dunbar ( 1998), Grooming, Gossip, and the Evolution of Language , Cambridge, MA: Harvard University Press. Fitzsimons Gavan J., Lehmann Donald R. ( 2004), “Reactance to Recommendations: When Unsolicited Advice Yields Contrary Responses,” Marketing Science , 23 ( 1), 82– 94. Google Scholar CrossRef Search ADS Folkman Susan ( 1984), “Personal Control and Stress and Coping Processes: A Theoretical Analysis,” Journal of Personality and Social Psychology , 46 ( 4), 839– 52. Google Scholar CrossRef Search ADS Friesen Justin P., Kay Aaron C., Eibach Richard P., Galinsky Adam D. ( 2014), “Seeking Structure in Social Organization: Compensatory Control and the Psychological Advantages of Hierarchy,” Journal of Personality and Social Psychology , 106 ( 4), 590– 609. Google Scholar CrossRef Search ADS Frizell Sam ( 2014), “Top 10 Gadgets,” http://time.com/collection-post/3582115/top-10-gadgets-2014/. Gatignon Hubert, Robertson Thomas S. ( 1986), “An Exchange Theory Model of Interpersonal Communication,” Advances in Consumer Research , 13, 534– 8. Godes David B., Mayzlin Dina ( 2004), “Using Online Conversations to Study Word-of-Mouth Communication,” Marketing Science , 23 ( 4), 545– 60. Google Scholar CrossRef Search ADS Graumann Carl Friedrich, Moscovici Serge ( 1986), Changing Conceptions of Crowd Mind and Behavior , New York: Springer. Google Scholar CrossRef Search ADS Gray Heather M., Ishii Keiko, Ambady Nalini ( 2011), “Misery Loves Company: When Sadness Increases the Desire for Social Connectedness,” Personality and Social Psychology Bulletin , 37 ( 11), 1438– 48. Google Scholar CrossRef Search ADS Griffit William, Veitch Russell ( 1971), “Hot and Crowded: Influence of Population Density and Temperature on Interpersonal Affective Behavior,” Journal of Personality and Social Psychology , 17 ( 1), 92– 8. Google Scholar CrossRef Search ADS Hennig-Thurau Thorsten, Gwinner Kevin P., Walsh Gianfranco, Gremler Dwayne D. ( 2004), “Electronic Word-of-Mouth via Consumer-Opinion Platforms: What Motivates Consumers to Articulate Themselves on the Internet?” Journal of Interactive Marketing , 18 ( 1), 38– 52. Google Scholar CrossRef Search ADS Inesi M. Ena, Botti Simona, Dubois David, Rucker Derek D., Galinsky Adam ( 2011), “Power and Choice: Their Dynamic Interplay in Quenching the Thirst for Power,” Psychological Science , 22 ( 8), 1042– 8. Google Scholar CrossRef Search ADS Jacobs Allan, Appleyard Donald ( 1987), “Toward an Urban Design Manifesto,” Journal of the American Planning Association , 53 ( 1), 112– 20. Google Scholar CrossRef Search ADS Jansen Bernard J., Zhang Mimi, Sobel Kate, Chowdury Abdur ( 2009), “Twitter Power: Tweets as Electronic Word of Mouth,” Journal of the American Society for Information Science and Technology , 60 ( 11), 2169– 88. Google Scholar CrossRef Search ADS Kalb Laura S., Keating John P. ( 1981), “The Measurement of Perceived Crowding,” Personality and Social Psychology Bulletin , 7 ( 4), 650– 4. Google Scholar CrossRef Search ADS Kay Aaron C., Gaucher Danielle, McGregor Ian, Nash Kyle ( 2010), “Religious Belief as Compensatory Control,” Personality and Social Psychology Review , 14 ( 1), 37– 48. Google Scholar CrossRef Search ADS Kay Aaron C., Gaucher Danielle, Napier Jamie L., Callan Mitchell J., Laurin Kristin ( 2008), “God and the Government: Testing a Compensatory Control Mechanism for the Support of External Systems,” Journal of Personality and Social Psychology , 95 ( 1), 18– 35. Google Scholar CrossRef Search ADS Kay Aaron C., Whitson Jennifer, Gaucher Danielle, Galinsky Adam D. ( 2009), “Compensatory Control: In the Mind, in Our Institutions,” Current Directions in Psychological Science , 18 ( 5), 264– 8. Google Scholar CrossRef Search ADS Keller Ed, Libai Barak ( 2009), “A Holistic Approach to the Measurement of WOM,” paper presented at the ESOMAR Worldwide Media Measurement Conference, Stockholm. Kemp Simon ( 2017), “Digital in 2017: Global Overview,” https://wearesocial.com/special-reports/digital-in-2017-global-overview. Langer Ellen J. ( 1975), “The Illusion of Control,” Journal of Personality and Social Psychology , 32 ( 2), 311– 28. Google Scholar CrossRef Search ADS Levav Jonathan, Zhu Rui (Juliet) ( 2009), “Seeking Freedom through Variety,” Journal of Consumer Research , 36 ( 4), 600– 10. Google Scholar CrossRef Search ADS Liu Thomas J., Steele Claude M. ( 1986), “Attributional Analysis as Self-Affirmation,” Journal of Personality and Social Psychology , 51 ( 3), 531– 40. Google Scholar CrossRef Search ADS Maeng Ahreum, Tanner Robin J., Soman Dilip ( 2013), “Conservative When Crowded: Social Crowding and Consumer Choice,” Journal of Marketing Research , 50 ( 6), 739– 52. Google Scholar CrossRef Search ADS Maslow Abraham H. ( 1943), “A Theory of Human Motivation,” Psychological Review , 50 ( 4), 370– 96. Google Scholar CrossRef Search ADS Milgram Stanley ( 1970), “The Experience of Living in Cities,” Science , 167 ( 3924), 1461– 8. Google Scholar CrossRef Search ADS Minube ( 2012), “Reasons to Visit Lisbon,” https://www.minube.net/travel/portugal/lisbon/lisbon. Mittal Chiraag, Griskevicius Vladas ( 2014), “Sense of Control under Uncertainty Depends on People’s Childhood Environment: A Life History Theory Approach,” Journal of Personality and Social Psychology , 107 ( 4), 621– 37. Google Scholar CrossRef Search ADS Novelli David, Drury John, Reicher Stephen, Stott Clifford ( 2013), “Crowdedness Mediates the Effect of Social Identification on Positive Emotion in a Crowd: A Survey of Two Crowd Events,” PLoS One , 8 ( 11), e78983. Google Scholar CrossRef Search ADS Nyer Prashanth U. ( 1997), “A Study of the Relationship between Cognitive Appraisals and Consumption Emotions,” Journal of the Academy of Marketing Science , 25 ( 4), 296– 304. Google Scholar CrossRef Search ADS Oldham Greg R., Rotchford Nancy L. ( 1983), “Relationships between Office Characteristics and Employee Reactions: A Study of the Physical Environment,” Administrative Science Quarterly , 28 ( 4), 542– 56. Google Scholar CrossRef Search ADS Paulus Paul B., Annis Angela B., Seta John J., Schkade Janette K., Matthews Robert W. ( 1976), “Density Does Affect Task Performance,” Journal of Personality and Social Psychology , 34 ( 2), 248– 53. Google Scholar CrossRef Search ADS Paulus Paul B., Matthews Robert W. ( 1980), “When Density Affects Crowd Performance,” Personality and Social Psychology Bulletin , 6 ( 1), 119– 24. Google Scholar CrossRef Search ADS Peluso Alessandro M., Bonezzi Andrea, De Angelis Matteo, Rucker Derek D. ( 2017), “Compensatory Word of Mouth: Advice as a Device to Restore Control,” International Journal of Research in Marketing , 34 ( 2), 499– 515. Google Scholar CrossRef Search ADS Qiu Simin ( 2015), “Water-Saving Tap Makes Pretty Patterns,” http://www.iflscience.com/technology/water-saving-tap-makes-pretty-patterns/. Rodin Judith ( 1976), “Density, Perceived Choice, and Response to Controllable and Uncontrollable Outcomes,” Journal of Experimental and Social Psychology , 12 ( 6), 564– 78. Google Scholar CrossRef Search ADS Rompay Thomas J. L., Galetzka Mirjam, Pruyn Ad T. H., Garcia Jaime Moreno ( 2008), “Human and Spatial Dimensions of Retail Density: Revisiting the Role of Perceived Control,” Psychology and Marketing , 25 ( 4), 319– 35. Google Scholar CrossRef Search ADS Rosenberg Morris ( 1989), Society and the Adolescent Self-Image , Middletown, CT: Wesleyan University Press. Rothbaum Fred, Weisz John R., Snyder Samuel S. ( 1982), “Changing the World and Changing the Self: A Two-Process Model of Perceived Control,” Journal of Personality and Social Psychology , 42 ( 1), 5– 37. Google Scholar CrossRef Search ADS Sedikides Constantine ( 1993), “Assessment, Enhancement, and Verification Determinants of the Self-Evaluation Process,” Journal of Personality and Social Psychology , 65 ( 2), 317– 38. Google Scholar CrossRef Search ADS Shankar Venkatesh, Venkatesh Alladi, Hofacker Charles, Naik Prasad ( 2010), “Mobile Marketing in the Retailing Environment: Current Insights and Future Research Avenues,” Journal of Interactive Marketing , 24 ( 2), 111– 20. Google Scholar CrossRef Search ADS Sherman David K., Cohen Geoffrey L. ( 2006), “The Psychology of Self-Defense: Self-Affirmation Theory,” Advances in Experimental Social Psychology , 38, 183– 242. Google Scholar CrossRef Search ADS Sherrod Drury R., Cohen Sheldon ( 1978), “Density, Perceived Control, and Design,” in Residential Crowding and Design , ed. Aiello John R., Baum Andrew, New York: Plenum Press, 217– 27. Sirgy M. Joseph ( 1982), “Self-Concept in Consumer Behavior: A Critical Review,” Journal of Consumer Research , 9 ( 3), 287– 300. Google Scholar CrossRef Search ADS Skinner Ellen A. ( 1996), “A Guide to Constructs of Control,” Journal of Personality and Social Psychology , 71 ( 3), 549– 70. Google Scholar CrossRef Search ADS Sommer Kristin L., Bourgeois Martin J. ( 2010), “Linking the Perceived Ability to Influence Others to Subjective Well-Being: A Need-Based Approach,” Social Influence , 5 ( 3), 220– 44. Google Scholar CrossRef Search ADS Spielberger Charles D., Gorsuch Richard L., Lushene Robert, Vagg P. R., Jacobs Gerard A. ( 1983), Manual for the State-Trait Anxiety Inventory , Palo Alto, CA: Consulting Psychologists Press. Staub Ervin, Tursky Bernard, Schwartz Gary E. ( 1971), “Self-Control and Predictability: Their Effects on Reactions to Aversive Stimulation,” Journal of Personality and Social Psychology , 18 ( 2), 157– 62. Google Scholar CrossRef Search ADS Steblay Nancy M. ( 1987), “Helping Behavior in Rural and Urban Environments: A Meta-Analysis,” Psychological Bulletin , 102 ( 3), 346– 56. Google Scholar CrossRef Search ADS Sundaram Suresh D., Mitra Kaushik, Webster Cynthia ( 1998), “Word-of-Mouth Communications: A Motivational Analysis,” Advances in Consumer Research , 25 ( 1), 527– 31. Thayer Robert E. ( 1989), The Biopsychology of Mood and Arousal , New York: Oxford University Press. Watson David, Clark Lee A., Tellegen Auke ( 1988), “Development and Validation of Brief Measures of Positive and Negative Affect: The PANAS Scales,” Journal of Personality and Social Psychology , 54 ( 6), 1063– 70. Google Scholar CrossRef Search ADS Wegner Daniel M., Wheatley Thalia ( 1999), “Apparent Mental Causation: Sources of the Experience of Will,” American Psychologist , 54 ( 7), 480– 92. Google Scholar CrossRef Search ADS Westbrook Robert A. ( 1987), “Product/Consumption-Based Affective Responses and Postpurchase Process,” Journal of Marketing Research , 24 ( 3), 258– 70. Google Scholar CrossRef Search ADS Whitson Jennifer A., Galinsky Adam D. ( 2008), “Lacking Control Increases Illusory Pattern Perception,” Science , 322 ( 5898), 115– 7. Google Scholar CrossRef Search ADS Wirth Louis ( 1938), “Urbanism as a Way of Life,” American Journal of Sociology , 44 ( 1), 1– 24. Google Scholar CrossRef Search ADS Wolny Julia, Mueller Claudia ( 2013), “Analysis of Fashion Consumers’ Motives to Engage in Electronic Word-of-Mouth Communication through Social Media Platforms,” Journal of Marketing Management , 29 ( 5–6), 562– 83. Google Scholar CrossRef Search ADS You Ya, Vadakkepatt Gautham G., Joshi Amit M. ( 2015), “A Meta-Analysis of Electronic Word-of-Mouth Elasticity,” Journal of Marketing , 79 ( 2), 19– 39. Google Scholar CrossRef Search ADS Zimbardo Philip G. ( 1969), “The Human Choice: Individuation, Reason, and Order versus Deindividuation, Impulse, and Chaos,” in Nebraska Symposium on Motivation , ed. Arnold William J., Levie David, Lincoln: University of Nebraska Press, 237– 307. © The Author(s) 2018. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc. All rights reserved. For Permissions, please e-mail: firstname.lastname@example.org This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/lega l/notices)
Journal of Consumer Research – Oxford University Press
Published: Feb 21, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera