3

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=hmcs20

Mass Communication and Society

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/hmcs20

Fear of Isolation and Perceived Affordances: The Spiral of Silence on Social Networking Sites Regarding Police Discrimination

Jesse Fox & Lanier Frush Holt

To cite this article: Jesse Fox & Lanier Frush Holt (2018) Fear of Isolation and Perceived Affordances: The Spiral of Silence on Social Networking Sites Regarding Police Discrimination, Mass Communication and Society, 21:5, 533-554, DOI: 10.1080/15205436.2018.1442480

To link to this article: https://doi.org/10.1080/15205436.2018.1442480

Published online: 13 Mar 2018.

Submit your article to this journal

Article views: 3054

View related articles

View Crossmark data

Citing articles: 22 View citing articles

Fear of Isolation and Perceived Affordances: The Spiral of Silence on Social Networking

Sites Regarding Police Discrimination

Jesse Fox and Lanier Frush Holt School of Communication The Ohio State University

The spiral of silence (SoS) framework elaborates the factors that determine whether individuals are willing to express their opinions in public. Although previous scholarship has examined differences in between face-to-face and computer- mediated communication, research studies have rarely tested how perceived affor- dances of the channel influence whether individuals express opinions or self-censor. In this study (N = 399), we examine several propositions of SoS within the context of discussing police discrimination on Facebook. To extend the theory’s relevance to social networking sites, we examined how users’ perceptions of network asso- ciation, social presence, anonymity, and persistence related to opinion expression. Findings indicate support for some of the theory’s original tenets, as well as the role of multiple perceived affordances in determining whether people will express an opinion to their online social network. We discuss the implications for measuring and understanding political expression and silencing on social media as well as offline.

Jesse Fox (Ph.D., StanfordUniversity, 2010) is an associate professor in the School ofCommunication at The Ohio State University. Her research interests include the effects of communication technologies, perceived affordances of technology, and intergroup communicaton online.

Lanier Frush Holt (Ph.D., Indiana University, 2010) is an assistant professor in the School of Communication at The Ohio State University. His research interests include race and media effects, specifically examining the effects media messages have on perceptions of Blacks and other traditionally marginalized groups.

Correspondence should be addressed to Jesse Fox, School of Communication, The Ohio State University, 154 North Oval Mall, 3016 Derby Hall, Columbus, OH 43210-1339. E-mail: [email protected]

Mass Communication and Society, 21:533–554, 2018 Copyright © Mass Communication & Society Division of the Association for Education in Journalism and Mass Communication ISSN: 1520-5436 print / 1532-7825 online DOI: https://doi.org/10.1080/15205436.2018.1442480

533

On August 9, 2014, Officer Darren Wilson shot and killed Mike Brown on the streets of Ferguson, Missouri, leading to civil unrest and widespread news coverage. Despite newfound attention to the issue, police violence against Black men and boys con- tinued: Chicago officer Jason Van Dyke fired 16 shots into teenager Laquan McDonald, several of which struck him as he lay on the ground dying; Officer Timothy Loehmann shot 12-year-old Tamir Rice without warning at a public recrea- tional center in Cleveland; and in broad daylight, Minnesota officer Jeronimo Yanez shot Philandro Castille multiple times as his girlfriend and her 4-year-old daughter watched in horror. These incidents, emblazoned across the front pages of newspapers and frequently broadcast on 24-hour news networks, have brought issues of race and police brutality to the forefront of the public’s consciousness.

Discussion of these issues has not been confined to traditional media; social media have become popular sites for information sharing and political discourse (Papacharissi, 2015). For example, a recent analysis of the Black Lives Matter movement identified more than 40 million mentions of the #BlackLivesMatter hashtag and related content on Twitter generated by more than 4.4 million users (Freelon, McIlwain, & Clark, 2016).

In an effort to better understand people’s willingness to express opinions or remain silent about police discrimination on social media, we adopt the spiral of silence (SoS) framework (Noelle-Neumann, 1974, 1993). We define police discrimination as citi- zens being treated differently by police because of their race, including racial profiling and the disproportionate number of Black individuals targeted by police brutality and shootings. The SoS framework has been adopted to examine race-related political issues because such topics are often sensitive and opinions can be polarized (Jeffres, Neuendorf, &Atkin, 1999;Moy, Domke,& Stamm, 2001). Opinions regarding police discrimination in the United States have been shown to be divisive (Pew Research Center, 2017). Further, consistent with the tenets of the SoS, there is a moral tension underlying the issue of police discrimination. Police are typically upheld as the arbiters of right and wrong; as such, policing efforts are often framed as moral actions (Cohen & Feldberg, 1991). Police discrimination is also at its heart a moral issue about racial inequality, dehumanization, and a quest for fair treatment by marginalized individuals (Prager, 2017); indeed, President Bill Clinton (1999) referred to racial profiling by police as a “morally indefensible” practice.

In addition to examining the SoS in this context, this study also aims to extend this theoretical framework to consider the nuances of speaking out on a social networking site (SNS). As Metzger (2009) noted, there are several features of computer-mediated environments that add complexity to the claims and assumptions of the SoS theory. For example, anonymity may make individuals more comfortable speaking out. On the other hand, the social connections enabled among different audiences may make people more likely to self-censor political expression, as they may be more uncertain about the majority opinion within these networks. Although previous studies have examined SoS

534 FOX AND HOLT

phenomena online (e.g., Gearhart & Zhang, 2015; Hampton et al., 2014; Ho & McLeod, 2008; McDevitt, Kiousis, & Wahl-Jorgensen, 2003; Stoycheff, 2016; Yun & Park, 2011), few have delved into the specific features of SNSs that may make them more or less similar to face-to-face interactions. Observed differences are chalked up to assumptions about the features of online environments without any direct measurement of these constructs. This practice is particularly proble- matic given the growing diversity of computer-mediated channels.

Our study addresses this gap by directly measuring the perceived affordances of these environments and examining how they may influence SoS processes on SNSs. First, we examine the claims of the SoS theory and outline the original tenets that are tested within this study. Next, we discuss various explanations and findings suggesting that the SoS may operate differently in computer-mediated communication (CMC) and SNS environments. Then, we argue for a perceived affordance approach to illuminate the specific aspects of online contexts that may influence the SoS. Our method tests both the original theoretical claims and the theoretical extensions offered by the perceived affordance approach.

SPIRAL OF SILENCE

According to the SoS (Noelle-Neumann, 1974, 1993), individuals are hesitant to voice their opinions regarding a political issue when they believe they are in the minority, because they fear social isolation. People who espouse the majority opinion feel comfortable expressing it publicly. Such expression, through both interpersonal and mass media outlets (e.g., radio and television talk shows, editorial content), builds the apparent scope and consonance of the majority opinion (Noelle-Neumann, 1974). An SoS ensues as individuals who believe they are in the minority avoid voicing their opinion due to the fear of social isolation, further reinforcing the majority opinion through an apparent lack of dissent. Because individuals wish to avoid social isolation, they must continuously scan the environment and use a “quasi-statistical sense” to determine public opinion on the issue (Noelle-Neumann, 1974, 1993).

Meta-analyses and critical syntheses have found limited support for the major theoretical tenets as well as Noelle-Neumann’s (1974, 1993) assertion of power- ful effects ranging from the individual’s willingness to speak out to the broader scope of shifting public opinions, voting behavior, and resultant political change (e.g., Glynn, Hayes, & Shanahan, 1997; Glynn & Huge, 2014; Scheufele & Moy, 2000; Shanahan, Glynn, & Hayes, 2007). Moreover, the scope of the theory is nearly impossible to test given its claims about the nature of mass media and the motivations of journalists; the imprecision about “public” contexts; and the need to make repeated assessments with the same participants over the course of several months, if not years (Matthes & Hayes, 2014; Shanahan et al., 2007). Some critics have suggested that although SoS has offered considerable heuristic

SPIRAL OF SILENCE ON SNSS 535

value, it is perhaps a better approach to focus on particular theoretical predictions relevant to a specific context instead of attempting to test the theory holistically (Hayes & Matthes, 2014; Matthes & Hayes, 2014; Salmon & Glynn, 1996). For example, scholars recommend examining the consonance of media coverage separately from individuals’ fear of isolation and willingness to speak out. Further, it is important to clarify whether individuals are willing to speak out in interpersonal, small-group, or mass-mediated contexts (e.g., being interviewed on television; Hayes & Matthes, 2014; Matthes & Hayes, 2014; Salmon & Glynn, 1996). In this way, new theorizing can emerge to address inconsistencies in findings and illuminate political expression and self-censorship across differ- ent contexts. Thus, our first goal in this study is to examine the primary SoS tenets directly related to individuals’ willingness to express opinions to others within the context of Facebook.

Across several elaborations and defenses of the theory, Noelle-Neumann (1974, 1993) stated that the issue under examination must be controversial or with a moral component. For this reason, we elected to investigate perceptions of police discrimination, which has emerged as a contentious and polarizing issue (Pew Research Center, 2017). Moreover, the conversation centers around race and equal treatment, which evokes a sense of morality (Moy et al., 2001; Prager, 2017). In conversation, a controversial moral issue is a risky topic; expressing disagreement with one’s audience could yield a negative reaction. Thus, it makes sense that those who fear alienation would be less likely to vocalize their opinions. Indeed, this is perhaps the most well-supported tenet of SoS theory (e.g., Glynn & Park, 1997; Hayes, Matthes, & Eveland, 2011; Kim, 2012; Matthes et al., 2012; Moy et al., 2001; Neuwirth, Frederick, & Mayo, 2007; Willnat, Lee, & Detenber, 2002). Thus, we hypothesize that a greater trait fear of isolation will be associated with a greater willingness to self-censor (H1).

The SoS theory noted, however, that some individuals will always be willing to express their opinion. These “hard-core” individuals have strong attitudes and are the passionate advocates and derogators of an issue (Noelle-Neumann, 1993). We expect that having stronger attitudes regarding police discrimination will be associated with less willingness to self-censor (H2).

Another primary tenet of SoS is that individuals speak out based on their assessment of the opinion climate and whether their opinion aligns with the majority (Noelle-Neumann, 1974, 1993). In two meta-analyses, Glynn and colleagues found weak but consistent support for the idea that people are more willing to speak out when they believe others share their views (Glynn et al., 1997; Glynn & Huge, 2014). Thus, we expect that perceptions of current opinion congruency will be associated with opinion expression such that those who believe their opinion aligns with U.S. public opinion will be less likely to self- censor (H3).

536 FOX AND HOLT

SoS also specifies that individuals may be more willing to speak out if they feel the tide is turning in their favor (Noelle-Neumann, 1974, 1993). Some studies have found support that perceptions of future agreement predict will- ingness to speak out even when the current opinion climate is not a factor (Gonzenbach, King, & Jablonski, 1999; Petrič & Pinter, 2002; Scheufele, Shanahan, & Lee, 2001). As such, we hypothesize that perceptions of future opinion congruency will be associated with opinion expression such that those who believe their opinion will align with anticipated public opinion in the future will be less likely to self-censor (H4).

SPIRAL OF SILENCE AND SELF-CENSORING IN ONLINE CONTEXTS

With the popularization of online interaction, scholars began to consider how the predictions of the SoS may hold up on the Internet. Many CMC channels are fundamentally different than face-to-face due to limited nonverbal cues, asyn- chronous communication, and greater control over whether and what one chooses to communicate (Lea & Spears, 1991; Rössler & Schulz, 2014). Given these differences, scholars have suggested that online environments may have a liberating effect in which users feel they may express opinions more freely (McDevitt et al., 2003; Metzger, 2009; Rössler & Schulz, 2014). Indeed, some early studies found that individuals were more likely to express opinions in chat rooms and online forums compared to face-to-face settings (Cortese & Seo, 2012; Ho & McLeod, 2008).

Subsequent studies comparing SNSs to face-to-face environments have not replicated this finding, however. Both Hampton et al. (2014) and Neubaum and Krämer (2018) found that people were less likely to speak up on an SNS than in a face-to-face context. Given these differences, it is important to investigate how these online environments vary and how these differences may influence the manifestation of the SoS.

Social Networking Sites as Opinion Climates

SNSs are an interesting area to examine the claims of the SoS given their status as masspersonal media. The original SoS was concerned with indivi- duals’ willingness to express their opinion “publicly” (Noelle-Neumann, 1974, 1993), yet many of the original studies adopted the “train test” paradigm in which “public” expression was operationalized as whether one would disclose one’s opinion to a stranger during a train ride. The “public” nature of this opinion expression, and this operationalization specifically, has been criticized by several scholars (Neuwirth et al., 2007). Hayes and Matthes (2014) speci- fically called for research examining “how individuated or personally

SPIRAL OF SILENCE ON SNSS 537

identifiable the person is when engaging in the behavior, the size of the audience witnessing the action, and the ability of the audience to administer sanction or disapproval” (p. 772). The masspersonal context of Facebook is an optimal site to examine these facets. It diverges from the train test given one’s opinion is immediately broadcast to a comparably large and more diverse audience but also differs from the media interview test because the audience consists of one’s social ties.

Given these differences in audience, we feel it is crucial in tests of SoS to consider how specific features of the audience and context may influence individuals’ willingness to speak out. The masspersonal context of an SNS seems to have the additional potential to evoke concerns of ostracism beyond general fears. Moreover, one may feel secure in voicing opinions on a variety of issues, but police discrimination is at its heart a racial issue. Many people avoid discussing race openly because it may be perceived as potentially threatening to others’ identities or there may be a higher fear of offending people (Jeffres et al., 1999; Lowery, Hardin, & Sinclair, 2001).

Although we agree that it is important to examine trait fear of isolation as prescribed by Glynn and McLeod (1985) and Hayes et al. (2011), Neuwirth and colleagues (2007) argued that the fear of isolation should be considered “a negative emotional state associated with the prospect of voicing one’s opinion about a given topic” that is tied to “a delimited sphere of discourse” (p. 452). Specific to police discrimination, some scholars have argued that individuals may feel they can more freely express themselves via social media than other channels (Bonilla & Rosa, 2015; Freelon et al., 2016). We also agree that it is important to factor in these context-specific elements of topic and audience, and it seems that including both types of fear can only expand the predictive power of a test of SoS. Thus, in addition to our hypothesis based on trait fear of isolation (see H1), we anticipate that higher context-specific fear of isolation (fear of voicing opinions about police discrimination on Facebook) will be associated with a greater willingness to self-censor (H5).

The context and content of Facebook is governed by the decisions of users and algorithms (Pariser, 2011); as such, users may perceive differing opinion climates (Schulz & Roessler, 2012). Findings are mixed in regards to the potential political heterogeneity of users’ SNS networks. Some studies suggest that individuals maintain ties with politically dissimilar others and engage in cross-cutting political discussion on SNSs (Bakshy, Messing, & Adamic, 2015; Heatherly, Lu, & Lee, 2017). If this is the case, users may perceive their Facebook climate as diverse and not necessarily aligned with their own opinions. Other findings suggest that when people are exposed to hostility toward their political views, users winnow their newsfeeds or online networks and actively block or defriend the offenders (Fox & Warber, 2015; John & Dvir-Gvirsman, 2015). This process may further narrow the subjectively perceived opinion

538 FOX AND HOLT

climate on Facebook and create a consonant environment or echo chamber that reflects one’s own opinions.

Regardless, these mixed findings suggest that it may be important to factor in the perceived opinion climate of one’s Facebook network when considering whether a user is willing to express political opinions on the site. Although Noelle-Neumann (1974, 1993) originally suggested that individuals are con- cerned with public opinion at a national level, others have argued that the perceived opinion climate of reference groups or specific audiences hold more predictive power regarding opinion expression (Glynn & McLeod, 1985; Glynn & Park, 1997; Price & Allen, 1990; Salmon & Kline, 1985). Regarding online contexts, two studies found that the online opinion climate predicted opinion expression, whereas broader public opinion did not (Nekmat & Gonzenbach, 2013; Yun & Park, 2011). Thus, we hypothesize that those who believe their opinion is congruent with the majority of their Facebook network will be less likely to self-censor (H6).

The SoS also accounts for perceptions of future change in the opinion climate. Although SNSs may mimic broader changes of public opinion, there are also individual factors that may predict shifts or stagnation among one’s Facebook network. For example, if a person maintains connections to a large contingent of conservative family members, that person may perceive minimal changes to the Facebook opinion climate even if public opinion is shifting. Alternatively, life events, such as the intent to relocate to a different state, may lead users to anticipate a shift in their Facebook opinion climate but not the public opinion. Thus, we expect that perceptions of future opinion congruency will be associated with opinion expression such that those who believe their opinion will align with the majority of their Facebook network in the future will be less likely to self- censor (H7).

Affordances of SNSs

It is important to examine SNSs as a specific context for the SoS. Although SNSs have many similarities with the CMC channels that have been commonly studied in SoS research, they also have distinct features that are relevant to the theory’s proposed mechanisms. Affordances are the properties of a technology that enable specific actions (Gibson, 1979), and how people perceive these affordances influences their behaviors (Norman, 1990). Although previous studies have attempted to manipulate these affordances (e.g., anonymity; Ho & McLeod, 2008; Yun & Park, 2011) or attribute their findings to them (e.g., social presence; Ho & McLeod, 2008; Kim, Kim, & Oh, 2014; McDevitt et al., 2003; Rössler & Schulz, 2014), researchers generally do not directly assess how users perceive them. In addition, given that there are often discrepancies between the manipula- tion of affordances and users’ perceptions (Fox & McEwan, 2017; Tao & Bucy,

SPIRAL OF SILENCE ON SNSS 539

2007), how users perceive site affordances may be the key to understanding what drives the SoS within and across online contexts.

A distinguishing characteristic of SNSs is that they afford network associa- tion, or the ability for people to identify friends of friends and other distal ties through visibly linked user profiles (boyd & Ellison, 2007; Fox & McEwan, 2017; Treem & Leonardi, 2013). On Facebook, people typically use their real names, and many “friends” on the site are people whom users know offline as well (boyd & Ellison, 2007); still, there is considerable variance among whom users are willing to friend (e.g., family members, coworkers, people from one’s past, or strangers) and how many “friends” they maintain (Vitak, 2012). Because SNSs enable geographically, socially, and politically disparate individuals to converge into a single opinion forum, this network association may assemble a rather heterogeneous audience. This heterogeneity, combined with the difficulty of keeping different types of ties separated on SNSs, can result in context collapse (Vitak, 2012). Users can feel uncomfortable when they are unable to maintain boundaries, and this discomfort may extend to groups with different political opinions. Thus, users who perceive high levels of network association may be less willing to express their opinions about a controversial issue given the ability for both strong and weak ties, as well as those who agree and disagree, to engage with a posted opinion (H8).

Network association allows others to view and interact on SNSs, which likely leads to a certain degree of social presence, or feelings that others are there sharing the same space (Short, Williams, & Christie, 1976). It is presumed that the more social presence one experiences in a mediated chan- nel, the more that communicative outcomes tend to reflect what is observed in face-to-face interactions (Short et al., 1976). Several scholars have argued that individuals will be more likely to express opinions online compared to face- to-face because of diminished social presence (e.g., Ho & McLeod, 2008; Jeffres et al., 1999; Kim et al., 2014; Metzger, 2009), although limited research has actually tested this assumption and measured users’ perceptions. One exception (Cortese & Seo, 2012) found that face-to-face environments evoked more social presence than CMC chat and that the experience of social presence was associated with less opinion expression. Following these find- ings, we expect that higher levels of social presence will be associated with a greater willingness to self-censor (H9).

One of the most commonly studied affordances in CMC research is anonym- ity. Because individuals can mask their identities when communicating online or feel that their online self-expression is not tied to their “real-world” identities, several theories suggest that individuals become deindividuated and engage in more disinhibited behavior in CMC compared to face-to-face (e.g., Lea & Spears, 1991; Suler, 2004). As Hayes and Matthes (2014) noted, an important clarification for the SoS is to assess how identifiable individuals feel when they

540 FOX AND HOLT

are expressing their opinions in a particular context. Although Yun and Park (2011) manipulated anonymity in an online test of SoS, they found no differ- ences across conditions; however, they acknowledged that the null finding may be attributable to not measuring whether participants actually felt anonymous during the interaction.

Indeed, it is crucial to assess perceived anonymity, or the degree to which users feel identifiable by a specific audience, regardless of the existence of explicit identity cues (Rössler & Schulz, 2014). Users may feel anonymous online even when they are using their real names because they still feel depersonalized, or as if their online self is separate from their offline existence (Suler, 2004). Given that perceptions of anonymity have been associated with more open expression in many online contexts (Suler, 2004), we still expect that lower perceptions of anonymity will be associated with a greater willingness to self-censor (H10).

A final affordance of CMC is persistence (Fox & McEwan, 2017; Treem & Leonardi, 2013). Compared to face-to-face interactions, where communication is usually ephemeral, online posts and comments are recorded and may last long after the original interaction transpires. Knowing that one’s political opinions might linger on the Internet for others to find may deter people from expressing an opinion (Stoycheff, 2016), particularly if they anticipate that attitudes may change in the future. Thus, we expect that higher perceptions of persistence will be associated with a greater willingness to self-censor (H11).

METHOD

Sample and Procedure

Study methods were approved by the Ohio State University Institutional Review Board in February 2016, and data collection commenced May 2016. U.S. Facebook users were recruited via Qualtrics for participation in an online survey about news and attitudes. Quota sampling was employed to obtain a roughly equal amount of Facebook users identifying as White and as Black. After providing consent, participants completed the survey, which lasted approximately 15 minutes. Fidelity checks were incorporated in the survey (e.g., among a randomized list of survey items, participants were instructed to “Select ‘Agree’ for this item”). If participants did not pass these checks, they were not included and sampling continued until the quota was reached.

Initially, 425 complete responses meeting these criteria were obtained. In examining the data, some participants did not have “friends” or social ties on Facebook (n = 7); given that this is a key aspect of the study, these participants were not included. An additional 19 participants provided incomplete or

SPIRAL OF SILENCE ON SNSS 541

suspicious data for open-ended items (e.g., entering the same digit across multi- ple items). Given that the quality of their data was suspect, they were removed from the data set.

The final sample (N = 399) included 207 women and 192 men (coded as 0 [woman] or 1 [man]) ranging in age from 18 to 84 (M = 45.68, SD = 17.11). Race was relatively evenly split; 195 participants identified as Black and 204 identified as White (coded as 0 [White] or 1 [Black]). Education level varied: 1 (did not complete high school), 4.3%; 2 (high school diploma), 21.1%; 3 (some college, trade school, or two-year degree), 42.6%; 4 (four-year college degree), 21.3%; and 5 (postgrad- uate degree), 10.8% (M = 3.13, SD = 1.00). Annual household income was assessed on a 9-point scale from 1 (less than $10,000) to 9 ($150,000 or more; M = 4.66, SD = 2.13). Following Noelle-Neumann’s (1974, 1993) recommenda- tions and previous research on racial issues (e.g., Moy et al., 2001; Willnat et al., 2002), these variables were included in the analysis.

Regarding Facebook, participants reported their number of friends on the site (M = 261.53, SD = 472.79; Mdn = 125). They also estimated the time they spend on Facebook on the average weekday and weekend day, and we used this to calculate their daily use (M = 135.49 minutes, SD = 129.36; Mdn = 95.71). Because distributions for these variables were not normal, they were log-trans- formed before inclusion in the analysis.

First, participants completed demographic and Facebook use measures. Then, they responded regarding their attitudes toward police discrimination. After parti- cipants reported on these attitudes, the following instructions were given: “Going forward, we will use the term ‘police discrimination’ to refer to unequal treatment of Blacks by cops, including racial profiling, the use of excessive force against Blacks (i.e., police brutality), and the shooting of Black suspects. When reading the following questions, please keep this definition in mind.” We provided these instructions to clarify our definition and ensure that participants were considering the scope of the concept when responding to the remaining questions regarding police discrimination. We also added instructions before relevant sets of items to remind people to consider the scope of police discrimination when responding.

Measures

Trait Fear of Isolation. Dispositional fear of social isolation was measured using the scale developed by Hayes et al. (2011). Participants responded to five items on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree), including “It is important to me to fit into the group I am with” and “It would bother me if no one wanted to be around me” (M = 3.91, SD = 1.49; Cronbach’s α = .87).

542 FOX AND HOLT

Contextual Fear of Isolation. Per Neuwirth et al. (2007), we derived a measure of context-specific fear of isolation considering both the issue and the audience. Participants responded to eight items from 1 (strongly disagree) to 7 (strongly agree), including “I am afraid that others will judge me if I express my opinion about police discrimination on Facebook” and “I feel I will lose friends or hurt relationships if I express my opinion about police discrimination on Facebook” (M = 4.51, SD = 1.40; α = .81).

Attitude Strength. For the purposes of this study, 10 Likert-style items were composed to assess participants’ perceptions regarding discriminatory practices by law enforcement. After using exploratory and confirmatory factor analyses to investigate the structure of these items, seven items were retained for use in the study. Cronbach’s alpha for the measure was α = .93. Participants originally responded on a 7-point scale: 1 (strongly disagree), 2 (disagree), 3 (slightly disagree), 4 (neither agree nor disagree), 5 (slightly agree), 6 (agree), and 7 (strongly agree); responses were recoded on a 4-point scale based on attitude strength: 0 (neither agree nor disagree) to 3 (strongly agree/ disagree). Items can be viewed in the appendix.

Perceived Opinion Congruency. Participants were asked to report their perceptions of what percentage of the United States believed that police discrimination was a problem (0%–100%). Following Ho and McLeod (2008), the attitude measure and the current perceived climate were recoded (support = 1; neutral/no opinion = 0; oppose = −1) and multiplied to yield a value indicating whether the participant’s attitudes were congruent (1) or incongruent (−1) with the current perceived public opinion. This value was also calculated considering what percentage of one’s Facebook network endorsed this opinion.

Then, participants were asked to report how likely they thought this opinion climate would change in the future, from 1 (very unlikely) to 5 (very likely). Following Ho and McLeod, the attitude measure and the future perceived climate were recoded (−1 = not changing, 0 = not sure, 1 = changing) and multiplied to yield a value indicating whether the participant’s attitudes were congruent (1), uncertain (0), or incongruent (−1) with future perceived public opinion. Again, this value was also calculated considering the likelihood of future change in one’s Facebook network.

Willingness to Self-Censor. To assess participants’ willingness to express their opinion on Facebook, we adapted the Willingness to Self-Censor Scale (Hayes, Glynn, & Shanahan, 2005b). Although the measure is often used as a trait predictor, Hayes et al. noted that it could be used as an outcome measure. For this study, we adapted the eight items, from 1 (strongly disagree) to 5 (strongly agree), to the specific context of Facebook. Items included (with

SPIRAL OF SILENCE ON SNSS 543

our additions in italics), “It is difficult for me to express my opinion on Facebook if I think others won’t agree with what I say” and “There have been many times when I have thought others on Facebook were wrong but I didn’t let them know” (M = 2.74, SD = .80). Cronbach’s alpha for the measure was .85.

Perceptions of SNS Affordances. Four subscales of the Perceived Affordances for Communication Channels Scale (Fox & McEwan, 2017) were implemented. Participants responded on 7-point scales from 1 (strongly disagree) to 7 (strongly agree). Confirmatory factor analysis was used to ensure factor structure. Four items measured network association (e.g., “Communicating on Facebook allows many members of my social network to be a part of my interaction”; M = 5.23, SD = 1.06; α = .78). Three items measured social presence (e.g., “Facebook makes it feel like other people are present”; M = 4.51, SD = 1.40; α = .85). Four items measured anonymity (e.g., “Facebook can mask my true identity when I’m posting”; M = 2.97, SD = 1.31; α = .79). Three items measured persistence (e.g., “Facebook keeps a record of my posts that people can go back and look at”; M = 5.63, SD = 1.01; α = .75).

RESULTS

Consistent with previous studies examining the SoS, hierarchical regression analysis was used to examine the model. No outliers were observed in the data set. Variance inflation factors were examined for indications of multicollinearity and found to be acceptable (all factors < 2).

In the first block, demographic variables and Facebook use variables were entered (race, gender, age, education, income, daily Facebook use, and number of Facebook friends). In the second block, individual perceptions and attitudes were entered (trait fear of isolation, context-specific fear of isolation, and attitude strength). In the third block, perceptions of congruence with current opinion as well as future opinion congruence were entered for both the United States and one’s Facebook network. The fourth block was composed of perceived affor- dances of Facebook (network association, social presence, anonymity, and persistence).

The unstandardized betas, standard errors, and standardized beta coefficients for the final model can be viewed in Table 1. In total, the R2 for the model was 42% (adjusted R2 = .39).

The first block predicted 8.5% of the variance in the model, F(7, 390) = 5.17, p < .001. Once all variables were entered in the model, only gender remained significant. Consistent with previous findings, women were more likely to self- censor than men.

544 FOX AND HOLT

The second block predicted 30.6% of the variance in the model, F(3, 387) = 64.83, p < .001. As predicted by H1 and H5, trait fear of isolation and context-specific fear of isolation were both positively associated with self-cen- soring. H2 was also supported; attitude strength was negatively associated with self-censoring.

The third block accounted for 0.4% of the variance in the model; the change was not significant, F(4, 383) = .70, p = .59. Only H4 was supported; con- gruency with future U.S. opinion was a significant predictor of self-censoring. Current U.S. opinion congruency (H3), current Facebook network opinion con- gruency (H6), and future Facebook network opinion congruency (H7) were not associated with self-censorship.

TABLE 1 Model to Predict Willingness to Self-Censor on Facebook

B SE β p

Constant 1.88 .32 Block 1

Race −.06 .07 −.04 .393 Gender −.18 .06 −.11 .005 Age .00 .00 −.07 .109 Education .02 .04 .03 .554 Income .01 .02 .03 .488 No. of friends −.03 .05 −.03 .541 Daily FB use −.08 .07 −.05 .243

R2 = .085 Block 2

Trait fear of isolation .06 .02 .12 .007 Contextual fear of isolation .35 .03 .49 < .001 Attitude strength −.15 .05 −.14 .001

ΔR2 = .306 Block 3

Current U.S. opinion congruency −.01 .04 −.01 .871 Current FB opinion congruency .01 .04 .01 .769 Future U.S. opinion congruency −.10 .05 −.10 .043 Future FB opinion congruency .09 .06 .08 .120

ΔR2 = .004 Block 4

Network association −.08 .04 −.11 .025 Social presence −.03 .03 −.05 .264 Anonymity .07 .03 .11 .010 Persistence .09 .04 .12 .010

ΔR2 = .025 Total R2 = .42

Note. FB = Facebook.

SPIRAL OF SILENCE ON SNSS 545

The fourth block predicted 2.5% of the variance in the model, F(4, 379) = 4.04, p = .003. Contrary to predictions, network association (H8) was negatively associated and anonymity (H10) was positively associated with self- censoring. Social presence was not associated with self-censorship, offering no support for H9. Supporting H11, persistence was associated with self-censorship.

DISCUSSION

Using the SoS framework, this study examined users’ willingness to self-censor their opinions about police discrimination on Facebook. Consistent with the theory’s original predictions, fear of isolation (measured as both a dispositional and context-specific variable) was associated with greater self-censorship, whereas having stronger attitudes was negatively associated with self-censorship. Opinion congruency with the current climate in the United States or one’s Facebook network did not predict silencing. In terms of future opinion con- gruency, congruency at the national level predicted silencing but not congruency with one’s Facebook network. Perceptions of three affordances (network asso- ciation, anonymity, and persistence) were associated with self-censorship, although only persistence was in the predicted direction.

One contribution of this study was the direct measurement and assessment of the perceived affordances of online environments that may influence the will- ingness for users to express an opinion or self-censor. Although a number of scholars have posited the role of affordances, manipulated them in experimental studies, or attributed findings to them, it is very rare for researchers examining the SoS to directly measure users’ perceptions. Clarifications of the theory have established that individual perceptions (whether of the opinion climate, the potential for isolation, or anticipated sanctions) must be assessed to determine whether an individual expresses an opinion. Similarly, our findings indicate that extending SoS online can be better understood by assessing individual percep- tions of the affordances of the online environment.

Consistent with Stoycheff’s (2016) speculation, we found that participants’ perceptions of the persistence of their posts predicted more self-censorship. The more they believed that their posts would linger online, the less willing they were to express their opinions about police discrimination. Users may fear that their posts may be discovered by others (such as future employers) or that the climate of opinion may shift in the future and leave them in the minority. Specific to our topic of study, people may be hesitant to criticize the police in case they need help from them or find themselves in trouble with the police in the future; speaking out on SNSs against authority figures or people in power may be seen as risky (Pearce & Vitak, 2016). Thus, persistence may be a key variable for understanding individuals’ willingness to speak out in online contexts.

546 FOX AND HOLT

Contrary to hypotheses, network association was negatively associated with self-censorship. Although we argued that perceptions of greater network associa- tion could yield a greater concern for blowback from the audience, users may actually feel that this association empowers them on Facebook because friends or likeminded individuals can either support the user or intervene and defend the user if needed. Similarly, although we were surprised to see anonymity asso- ciated with greater self-censoring, this finding makes sense from this perspective: If users feel identifiable rather than faceless in their network, they may expect their known ties to support or defend them and thus feel more comfortable expressing their opinions. To elucidate these affordances, future research should also consider the consequences users anticipate for expressing their opinions (Neubaum & Krämer, 2018; Rössler & Schulz, 2014).

In these ways, an SNS like Facebook is fundamentally different than many contexts traditionally tested using SoS theory and may challenge some of its assumptions. Because of the presence of known ties as well as the visibility and persistence of others’ opinions, a “quasi-statistical sense” may not be necessary to determine the opinion climate. Fear of social isolation may be mitigated on a site like Facebook because we are typically surrounded by our closest ties, who either share our opinions or love us in spite of them. For the same reason, users’ self-expression or self-censorship may not be motivated by their current or future opinion congruency with a numerical majority as SoS claims. Users may be more concerned with the congruency of their opinions with a certain quality of tie rather than a quantity of ties. For example, lesbian, gay, bisexual, transgender, queer individuals may self-censor about lesbian, gay, bisexual, transgender, questioning issues on Facebook if they are connected to key ties that disapprove (e.g., a homophobic parent) even when they feel the majority of their network is supportive (Fox & Warber, 2015). In our study, individuals may have refrained from discussing police discrimination even when they perceived high opinion congruency because they may not have wanted to upset a specific person in their network (e.g., a friend whose spouse is a police officer) or may have wanted to avoid triggering an unwelcome interaction (e.g., a response from a racist cousin). Thus, masspersonal, anchored SNSs like Facebook challenge some of the funda- mental premises of SoS theory.

Examining perceived affordances may help clarify several aspects of the SoS and why it may manifest differently across both channels and topics. Affordances vary so extensively in their implications, however, that it is unlikely that one simple “affordances” box can be drawn with designated paths into a model of the SoS. Some perceived affordances, such as anonymity or the opportunity for social feedback, may influence context-specific fear of isolation or willingness to self-censor. Anonymity may be particularly important for someone with high trait fear of social isolation, as they want the assurances that they will not be ostracized for expressing an opinion. Asynchronicity and editability may directly

SPIRAL OF SILENCE ON SNSS 547

influence the willingness to self-censor; if a user has time to research and carefully craft a political statement, that user may be more likely to make one than in an immediate, face-to-face environment.

Further, different combinations of these affordances may yield different out- comes. In our study, individuals generally perceived Facebook as high in asso- ciation but low in anonymity. Anonymity may matter less in an environment with high levels of association because those ties may be meaningful (e.g., “Internet friends”) even if your offline identity is unknown to them. Persistence may be less of a concern in a more anonymous environment because comments would not feel tied to one’s identity. Persistence may matter more in an environment high in association, as people may be concerned that known or more socially significant ties will dig through their previous posts. Future studies should compare self-censoring across mediated platforms to clarify how combi- nations of affordances are perceived.

A secondary contribution to this study was addressing a methodological issue in SoS framework by assessing both trait fear of isolation and context-specific fear of isolation. Although the two were correlated (r = .28, p < .001), they both independently contributed to opinion expression. Because we had conceptualized self-censoring as an outcome in this study, however, we did not examine trait willingness to self-censor alongside our context-specific assessment. Future research on the SoS, particularly in online settings where the audience varies considerably, should assess both trait and context-specific fear and willingness to self-censor.

In addition to our hypotheses, our final model also revealed the role of demographic characteristics. Supporting Noelle-Neumann’s original theorizing (as well as previous research with the Willingness to Self-Censor Scale by Hayes, Glynn, & Shanahan, 2005a; Hayes et al., 2005b), women were more likely to self-censor than men in our sample. In the final model, there were no differences in between White and Black participants’ willingness to self-censor. Future research should explore other variables that may predict when individuals are willing to speak out on a race-related issue, such as the strength of one’s ethnoracial identity, personal experience, attitudes toward outgroup members, or social dominance orientation.

Our approach for examining the SoS framework has inevitable limitations. First, we abided by the advice of several scholars and did not attempt to tackle the entire scope of the theory (Hayes & Matthes, 2014; Matthes & Hayes, 2014). We focused on the individual and social predictors of opinion expression rather than the role of the mass media. Given that Noelle-Neumann’s (1974) claims of a consonant mass media have been largely unsupported and theoretically chal- lenged (Glynn & McLeod, 1985; Salmon & Kline, 1985), particularly given the emergence of a fragmented online environment since the advent of the theory

548 FOX AND HOLT

(Metzger, 2009; Schulz & Roessler, 2012), we chose to examine the aspects of SoS that have garnered support.

Although some have argued that survey techniques are sufficient or even optimal for examining the predictions of SoS (Scheufele & Moy, 2000), others have suggested that other techniques, particularly experimental approaches, are more suitable (Matthes & Hayes, 2014; Salmon & Glynn, 1996). Although the SoS theorizes directionality, our use of survey techniques limits our ability to make causal claims.

In this study, we examined four affordances guided by previous SoS scholar- ship. Future research should investigate perceptions of other affordances. For example, Salmon and Glynn (1996) noted that “direct” or “immediate” feedback may influence whether people will express opinions in different contexts. Users often weigh the potential for social feedback before disclosing on SNSs (Fox & Moreland, 2015). Researchers could measure whether perceptions of affordances such as the speed or synchronicity of feedback, the personalization of feedback, or the bandwidth or cues transmitted in feedback influence the likelihood of expressing one’s opinions in various channels (Fox & McEwan, 2017). Equally important to consider would be constraints within these systems: Whereas affordances enable actions, constraints prevent actions from occurring. Some of these may be automatically built into the system; others may be enforced or triggered by other users or Facebook’s employees or algorithms. For example, Facebook users may be prevented from commenting on another user’s page if they are not “friends,” which prohibits the extent of expression. Facebook users can also flag material for offensiveness, which may lead to the removal of a post; thus, users are somewhat constrained in what they are permitted to express on the platform. Assessing both the costs and utility of a channel such as Facebook may also predict why individuals choose to express on Facebook or not (Trumbo, 1998).

Previous research has assessed several additional variables that may lend more insight into our findings. Issue knowledge (Salmon & Neuwirth, 1990), issue involvement (Kim et al., 2014; Lin & Pfau, 2007), and attitude certainty (Matthes, Morrison, & Schemer, 2010), for example, have been shown to predict a greater willingness to express opinions. Given that issue involvement is likely tied to race and social identity, issue involvement may have been particularly insightful to our study. We also did not assess the degree to which individuals felt that police discrimination is a controversial or morally charged issue, which may influence their willingness to speak out. Another potential shortcoming of our study is that we adapted a trait measure for use as a dependent variable. Although participants were instructed to consider police discrimination in the survey, we did not adapt individual items in the scale to mention police discrimination, which would have strengthened the validity of these items as specific to this context.

SPIRAL OF SILENCE ON SNSS 549

Additional contextual variables should also be considered. The argumentative- ness or hostility of the deliberative environment has been shown to quell opinion expression (Hayes, 2007) and may be particularly relevant to online contexts (Gearhart & Zhang, 2015). Outcomes aside from self-censorship should be exam- ined, such as expressing one’s opinion, “speaking out” versus “speaking up,” misrepresenting one’s viewpoint, or changing the subject (Gearhart & Zhang, 2015; Hayes, 2007; McDevitt et al., 2003; Neuwirth et al., 2007). SNSs also afford different methods of self-expression beyond direct statements. Future research could explore if there are differences in the willingness to express one’s opinion through explicit posts on one’s own profile as compared to “likes,” “upvotes,” and other aggregated approval cues; commenting on others’ posted opinions; or sharing links or others’ posts that express opinions.

Finally, we examined only one issue (police discrimination) and one context (U.S.-based Facebook) in our study. Many scholars have noted that results may be contingent on the nature of the issue (Salmon & Glynn, 1996) and studies that have observed multiple issues rarely show consistent findings across topics (e.g., Gearhart & Zhang, 2018; Willnat et al., 2002). It seems likely that various affordances may become more crucial regarding certain issues and in certain contexts. For example, perceptions of anonymity or the persistence of one’s posts may be far more predictive of opinion expression in countries with restricted speech or a history of imprisoning dissenters, particularly if the issue is sensitive or critical of the government (Stoycheff, 2016). Future research must directly assess users’ perceptions of these affordances rather than making assumptions that CMC is inherently anonymous.

The Internet continues to provide users with a variety of outlets for informa- tion sharing and opinion expression. The SoS framework has demonstrated its potential in various online settings, although scholars must continue to parse out the differences in perceived affordances and opinion climates across these sites to fully understand the nature of self-censorship and political expression online.

ACKNOWLEDGMENTS

The authors would like to thank Dr. Kelly Dillon for building the online survey and Dr. Chip Eveland for his feedback on an earlier draft of this paper.

REFERENCES

Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and opinion on Facebook. Science, 348(6239), 1130–1132. doi:10.1126/science.aaa1160

Bonilla, Y., & Rosa, J. (2015). #Ferguson: Digital protest, hashtag ethnography, and the racial politics of social media in the United States. American Ethnologist, 42, 4–17. doi:10.1111/amet.12112

550 FOX AND HOLT

boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, 210–230. doi:10.1111/j/1083-6101.2007.00393.x

Clinton, B. (1999, June 9). Remarks by the president on civil rights law enforcement. Washington, DC: Office of the Press Secretary. Retrieved from https://clintonwhitehouse2.archives.gov/WH/ New/html/19990609.html

Cohen, H. S., & Feldberg, M. (1991). Power and restraint: The moral dimension of police work. New York, NY: Praeger.

Cortese, J., & Seo, M. (2012). The role of social presence in opinion expression during FtF and CMC discussions. Communication Research Reports, 29, 44–53. doi:10.1080/08824096.2011.639913

Fox, J., & McEwan, B. (2017). Distinguishing technologies for social interaction: The perceived social affordances of communication channels scale. Communication Monographs, 84, 298–318. doi:10.1080/03637751.2017.1332418

Fox, J., & Moreland, J. J. (2015). The dark side of social networking sites: An exploration of the relational and psychological stressors associated with Facebook use and affordances. Computers in Human Behavior, 45, 168–176. doi:10.1016/j.chb.2014.11.083

Fox, J., & Warber, K. M. (2015). Queer identity management and political self-expression on social networking sites: A co-cultural approach to the spiral of silence. Journal of Communication, 65, 79–100. doi:10.1111/jcom.12137

Freelon, D., McIlwain, C. D., & Clark, M. D. (2016). Beyond the hashtags: #Ferguson, #Blacklivesmatter, and the online struggle for offline justice. Washington, DC: American University Center for Media & Social Impact.

Gearhart, S., & Zhang, W. (2015). “Was it something I said?” “No, it was something you posted!” A study of the spiral of silence theory in social media contexts. Cyberpsychology, Behavior, & Social Networking, 18, 208–213. doi:10.1089/cyber.2014.0443

Gearhart, S., & Zhang, W. (2018). Same spiral, different day? Testing the spiral of silence across issue types. Communication Research, 45, 34–54. doi:10.1177/0093650215616456

Gibson, J. J. (1979). The ecological approach to visual perception. Boston, MA: Houghlin Mifflin. Glynn, C. J., Hayes, A. F., & Shanahan, J. (1997). Perceived support for one’s opinions and the

willingness to speak out: A meta-analysis of survey studies on the “spiral of silence.” Public Opinion Quarterly, 61, 452–463. doi:10.1086/297808

Glynn, C. J., & Huge, M. E. (2014). Speaking in spirals: An updated meta-analysis on the spiral of silence. In W. Donsbach, C. T. Salmon, & Y. Tsfati (Eds.), The spiral of silence: New perspectives on communication and public opinion (pp. 65–72). New York, NY: Routledge.

Glynn, C. J., & McLeod, J. M. (1985). Implications of the spiral of silence theory for communication and public opinion research. In K. R. Sanders, L. L. Kaid, & D. Nimmo (Eds.), Political communication yearbook 1984 (pp. 43–65). Carbondale, IL: Southern Illinois University Press.

Glynn, C. J., & Park, E. (1997). Reference groups, opinion intensity, and public opinion expression. International Journal of Public Opinion Research, 9, 213–232. doi:10.1093/ijpor/9.3.213

Gonzenbach, W. J., King, C., & Jablonski, P. (1999). Homosexuals and the military: An analysis of the spiral of silence. Howard Journal of Communication, 10(4), 281–296. doi:10.1080/ 106461799246762

Hampton, K. N., Rainie, L., Lu, W., Dwyer, M., Shin, I., & Purcell, K. (2014). Social media and the ‘spiral of silence’. Washington, DC: Pew Research Center.

Hayes, A. (2007). Exploring the forms of self-censorship: On the spiral of silence and the use of opinion expression avoidance strategies. Journal of Communication, 57, 785–802. doi:10.1111/ j.1460-2466.2007.00368.x

Hayes, A. F., Glynn, C. J., & Shanahan, J. (2005a). Validating the willingness to self-censor scale: Individual differences in the effect of the climate of opinion on opinion expression. International Journal of Public Opinion Research, 17, 443–455. doi:10.1093/ijpor/edh072

SPIRAL OF SILENCE ON SNSS 551

Hayes, A. F., Glynn, C. J., & Shanahan, J. (2005b). Willingness to self-censor: A construct and measurement tool for public opinion research. International Journal of Public Opinion Research, 17, 298–323. doi:10.1093/ijpor/edh073

Hayes, A. F., & Matthes, J. (2014). Self-censorship, the spiral of silence, and contemporary political communication. In K. H. Jamieson & K. Kenski (Eds.), Oxford handbook on political commu- nication (pp. 763–776). Oxford, UK: Oxford University Press. doi:10.1093/oxfordhb/ 9780199793471.013.018

Hayes, A. F., Matthes, J., & Eveland, W. P. (2011). Stimulating the quasi-statistical organ: Fear of social isolation motivates the quest for knowledge of the opinion climate. Communication Research, 40, 439–462. doi:10.1177/0093650211428608

Heatherly, K. A., Lu, Y., & Lee, J. K. (2017). Filtering out the other side? Cross-cutting and like- minded discussions on social networking sites. New Media & Society, 19, 1271–1289. doi:10.1177/ 1461444816634677

Ho, S. S., & McLeod, D. M. (2008). Social-psychological influences on opinion expression in face- to-face and computer-mediated communication. Communication Research, 35, 190–207. doi:10.1177/0093650207313159

Jeffres, L. W., Neuendorf, K. A., & Atkin, D. (1999). Spirals of silence: Expressing opinions when the climate of opinion is unambiguous. Political Communication, 16, 115–131. doi:10.1080/ 105846099198686

John, N. A., & Dvir-Gvirsman, S. (2015). “I don’t like you any more”: Facebook unfriending by Israelis during the Israel–Gaza conflict of 2014. Journal of Communication, 65, 953–974. doi:10.1111/jcom.12188

Kim, S. H. (2012). Testing fear of isolation as a causal mechanism: Spiral of silence and genetically modified (GM) foods in South Korea. International Journal of Public Opinion Research, 24, 306– 324. doi:10.1093/ijpor/eds017

Kim, S. H., Kim, H., & Oh, S. H. (2014). Talking about genetically modified (GM) foods in South Korea: The role of the Internet in the spiral of silence process. Mass Communication & Society, 17, 713–732. doi:10.1080/15205436.2013.847460

Lea, M., & Spears, R. (1991). Computer-mediated communication, deindividuation, and group decision-making. International Journal of Man Machine Studies, 34, 283–301. doi:10.1016/ 0020-7373(91)90045-9

Lin, W. K., & Pfau, M. (2007). Can inoculation work against the spiral of silence? A study of public opinion on the future of Taiwan. International Journal of Public Opinion Research, 19, 155–172. doi:10.1093/ijpor/edl030

Lowery, B. S., Hardin, C. D., & Sinclair, S. (2001). Social influence effects on automatic racial prejudice. Journal of Personality & Social Psychology, 81, 842–855. doi:10.1037/0022-3514.81.5.842

Matthes, J., & Hayes, A. F. (2014). Methodological conundrums in spiral of silence research. In W. Donsbach, C. T. Salmon, & Y. Tsfati (Eds.), The spiral of silence: New perspectives on commu- nication and public opinion (pp. 54–55). New York, NY: Routledge.

Matthes, J., Hayes, A. F., Rojas, H., Shen, F., Min, S. J., & Dylko, I. B. (2012). Exemplifying a dispositional approach to cross-cultural spiral of silence research: Fear of social isolation and the inclination to self-censor. International Journal of Public Opinion Research, 24, 287–305. doi:10.1093/ijpor/eds015

Matthes, J., Morrison, K. R., & Schemer, C. (2010). A spiral of silence for some: Attitude certainty and the expression of political minority opinions. Communication Research, 37, 774–800. doi:10.1177/0093650210362685

McDevitt, M., Kiousis, S., & Wahl-Jorgensen, K. (2003). Spiral of moderation: Opinion expression in computer-mediated discussion. International Journal of Public Opinion Research, 15, 454–470. doi:10.1093/ijpor/15.4.454

552 FOX AND HOLT

Metzger,M. J. (2009). The study ofmedia effects in the era of Internet communication. InR. L.Nabi&M.B. Oliver (Eds.), The Sage handbook of media processes and effects (pp. 561–576). Los Angeles, CA: Sage.

Moy, P., Domke, D., & Stamm, K. (2001). The spiral of silence and public opinion on affirmative action. Journalism & Mass Communication Quarterly, 78, 7–25. doi:10.1177/107769900107800102

Nekmat, E., & Gonzenbach, W. J. (2013). Multiple opinion climates in online forums: Role of website source reference and within-forum opinion congruency. Journalism & Mass Communication Quarterly, 90, 736–756. doi:10.1177/1077699013503162

Neubaum, G., & Krämer, N. C. (2018). What do we fear? Expected sanctions for expressing minority opinions in offline and online communication. Communication Research, 45, 139–164. doi:10.1177/0093650215623837

Neuwirth, K., Frederick, E., & Mayo, C. (2007). The spiral of silence and fear of isolation. Journal of Communication, 57, 450–468. doi:10.1111/j.1460-2466.2007.00352.x

Noelle-Neumann, E. (1974). The spiral of silence: A theory of public opinion. Journal of Communication, 24, 43–51. doi:10.1111/j.1460-2466.1974.tb00367.x

Noelle-Neumann, E. (1993). The spiral of silence: Public opinion, our social skin (2nd ed.). Chicago, IL: University of Chicago Press.

Norman, D. A. (1990). The design of everyday things. New York, NY: Doubleday. Papacharissi, Z. (2015). Affective publics: Sentiment, technology, and politics. New York, NY: Oxford. Pariser, E. (2011). The filter bubble: How the new personalized web is changing what we read and

how we think. New York, NY: Penguin. Pearce, K. E., & Vitak, J. (2016). Performing honor online: The affordances of social media for

surveillance and impression management in an honor culture. New Media & Society, 18, 2595– 2612. doi:10.1177/1461444815600279

Petrič, G., & Pinter, A. (2002). From social perception to public expression of opinion: A structural equation modeling approach to the spiral of silence. International Journal of Public Opinion Research, 14, 37–53. doi:10.1093/ijpor/14.1.37

Pew Research Center. (2017, January 11). Behind the badge. Washington, DC: Author. Prager, J. (2017). Do Black lives matter? A psychoanalytic exploration of racism and American

resistance to reparations. Political Psychology, 38, 637–651. doi:10.1111/pops.12436 Price, V., & Allen, S. (1990). Opinion spirals, silent and otherwise: Applying small-group research to public

opinion phenomena. Communication Research, 17, 369–392. doi:10.1177/009365090017003005 Rössler, P., & Schulz, A. (2014). Public opinion expression in online environments. In W. Donsbach,

C. T. Salmon, & Y. Tsfati (Eds.), The spiral of silence: New perspectives on communication and public opinion (pp. 101–118). New York, NY: Routledge.

Salmon, C. T., & Glynn, C. J. (1996). Spiral of silence: Communication and public opinion as social control. In M. B. Salwen & D. W. Stacks (Eds.), An integrated approach to communication theory and research (pp. 165–180). Mahwah, NJ: Erlbaum.

Salmon, C. T., & Kline, F. G. (1985). The spiral of silence ten years later: An examination and evaluation. In K. R. Sanders, L. L. Kaid, & D. Nimmo (Eds.), Political communication yearbook 1984 (pp. 3–30). Carbondale, IL: Southern Illinois University Press.

Salmon, C. T., & Neuwirth, K. (1990). Perceptions of opinion “climates” and willingness to discuss the issue of abortion. Journalism & Mass Communication Quarterly, 67, 567–577. doi:10.1177/ 107769909006700312

Scheufele, D. A., & Moy, P. (2000). Twenty-five years of the spiral of silence: A conceptual review and empirical outlook. International Journal of Public Opinion Research, 12, 3–28. doi:10.1093/ijpor/12.1.3

Scheufele, D. A., Shanahan, J., & Lee, E. (2001). Real talk manipulating the dependent variable in spiral of silence research. Communication Research, 28, 304–324. doi:10.1177/009365001028003003

Schulz, A., & Roessler, P. (2012). The spiral of silence and the Internet: Selection of online content and the perception of the public opinion climate in computer-mediated communication environments. International Journal of Public Opinion Research, 24, 346–367. doi:10.1093/ijpor/eds022

SPIRAL OF SILENCE ON SNSS 553

Shanahan, J., Glynn, C., & Hayes, A. (2007). The spiral of silence: A meta-analysis and its impact. In R. W. Preiss, B. M. Gayle, N. Burrell, M. Allen, & J. Bryant (Eds.), Mass media effects: Advances through meta-analysis (pp. 415–427). Mahwah, NJ: Erlbaum.

Short, J., Williams, E., & Christie, B. (1976). The social psychology of telecommunications. New York, NY: Wiley.

Stoycheff, E. (2016). Under surveillance: Examining Facebook’s spiral of silence effects in the wake of NSA Internet monitoring. Journalism & Mass Communication Quarterly, 93, 296–311. doi:10.1177/1077699016630255

Suler, J. (2004). The online disinhibition effect. CyberPsychology & Behavior, 7, 321–326. doi:10.1089/1094931041291295

Tao, C. C., & Bucy, E. P. (2007). Conceptualizing media stimuli in experimental research: Psychological versus attribute-based definitions. Human Communication Research, 33, 397–426. doi:10.1111/j.1468-2958.2007.00305.x

Treem, J., & Leonardi, P. (2013). Social media use in organizations: Exploring the affordances of visibility, editability, persistence, and association. In C. T. Salmon (Ed.), Communication yearbook (Vol. 38, pp. 143–189). New York, NY: Routledge.

Trumbo, C. W. (1998). Communication channels and risk information: A cost-utility model. Science Communication, 20, 190–203. doi:10.1177/1075547098020002002

Vitak, J. (2012). The impact of context collapse and privacy on social network site disclosures. Journal of Broadcasting & Electronic Media, 56, 451–470. doi:10.1080/08838151.2012.732140

Willnat, L., Lee, W., & Detenber, B. H. (2002). Individual-level predictors of public outspokenness: A test of the spiral of silence theory in Singapore. International Journal of Public Opinion Research, 14, 391–412. doi:10.1093/ijpor/14.4.391

Yun, G. W., & Park, S. Y. (2011). Selective posting: Willingness to post a message online. Journal of Computer-Mediated Communication, 16, 201–227. doi:10.1111/j.1083-6101.2010.01533.x

APPENDIX

Items Assessing Attitudes Toward Police Discrimination

1. The police rely on racial profiling when enforcing the law (for example, when deciding which car to pull over).

2. If a White person and a Black person were under consideration for the same crime, the cops would be more likely to suspect the Black person.

3. The police are more likely to use excessive force if a suspect is Black. 4. Most members of law enforcement are racist against Blacks, even if they

don’t mean to be. 5. Most police would not hesitate to shoot a Black man, even if he was

unarmed. 6. Police brutality against Blacks is a serious problem. 7. Police officers who shoot unarmed Black men should be charged with a hate

crime.

554 FOX AND HOLT