Selective Exposure for Better or Worse: Its Mediating Role for Online News Impact on Political Participation

Journal of Computer-Mediated Communication Selective Exposure for Better or Worse: Its Mediating Role for Online News Impact on Political Participation Silvia Knobloch-Westerwick School of Communication, The Ohio State University, 3020 Derby Hall, 154 N. Oval Mall, Columbus, OH 43210 Benjamin K. Johnson School of Communication, The Ohio State University, 3016 Derby Hall, 154 N. Oval Mall, Columbus, OH 43210 The role of selective exposure in the relationship between online news use and political participation is examined. American adults (N = 205) completed a 2-session online study that measured political interest and online news use, unobtrusively observed selective exposure, and finally measured political participation likelihood. Online news use and selective exposure to attitude-consistent information were modeled as sequential mediators between political interest and participation likelihood. While greater political interest increased both participation likelihood and online news use, online news use ultimately depressed participation likelihood by reducing selective exposure to attitude-consistent news. The findings demonstrate that selective exposure is a fundamental process that must be considered when testing the effect of Internet use on political participation. Key words: selective exposure, Internet, participation, elections, news. doi:10.1111/jcc4.12036 The advent of the Internet has revolutionized access to political information and participation in public affairs, according to some scholars (Benkler, 2006; Bimber, 1998; Krueger, 2002; Weber, Loumakis, & Bergman, 2003). Clearly, the Internet provides unprecedented access to information and may lower barriers for many political activities such as contacting representatives and engaging in political speech. However, it also presents increased opportunities and means by which individuals may self-select into information and news that only reinforce existing views or into entertainment and amusement that bypass political concerns entirely. Such confirmation bias and selective avoidance have been said to impair democratic processes by increasing polarization and reducing deliberation (Sunstein, 2001). The net effect of the Internet on citizenry s involvement in the democratic process is thus a critical societal concern. Yet whether Internet use ultimately fosters or undermines political participation remains a topic of discussion among scholars. In the past 15 years, a possible link between online behavior and political participation has frequently been examined. A meta-analysis (Boulianne, 2009) of this body of research found that Internet use has generally been shown to have a positive, but very modest, effect on political and civic engagement, with online news use playing a specific role in this process. Meanwhile, a separate debate regarding effects of Accepted by previous editor Maria Bakardjieva 184 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

Internet use on political behavior has focused on selectivity in political information exposure (Bennett & Iyengar, 2008; Holbert, Garrett, & Gleason, 2010), drawing on much less empirical evidence. The present investigation aims to consider concepts from these different debates simultaneously, building from the framework of Noelle-Neumann s (1974) spiral of silence theory. A sequential mediation model (Taylor, MacKinnon, & Tein, 2008) is proposed, with political interest leading to online news use habit, followed by habit influencing selective exposure, and selectivity ultimately affecting political participation. The model is assessed with a two-session online study using a non-student adult sample. Internet Use and Political Participation Research examining political participation as a consequence of media use, especially Internet use, has generally operated under the assumption that increased access to a large, diverse set of political information may help reinvigorate civic life (Boulianne, 2009, p. 205). Her meta-analysis of this research demonstrated that Internet use was generally shown to have a positive but small effect on political participation. This link emerged as stronger when specifically examining online news use, but any association between use and engagement tended to vanish when controlling for political interest. Furthermore, random and nationally representative samples were even less likely to show any effect of online activity on engagement. Many of the 39 studies included in Boulianne s meta-analysis (2009) yielded non-significant findings or evidence contradicting the idea that Internet use enhances political participation. Furthermore, this literature on Internet effects on political engagement lacks a coherent theoretical framework, with uncertainty regarding the roles and casual direction of concepts (Boulianne, 2009). It also fails to consider the important role that perceptions of others political attitudes play in influencing participation. The present investigation addresses this gap by drawing on spiral of silence theory (Noelle-Neumann, 1974) to conceptualize the processes involved in Internet use effects on political participation. Spiral of Silence Theory Spiral of silence theory (Noelle-Neumann, 1974) suggests that individuals form an impression of the distribution of opinions based on interpersonal encounters and media messages. Furthermore, individuals who perceive their opinion group as being dominant or ascendant are more likely to express their opinion publicly, whereas those who perceive their opinions as the minority or in decline choose to censor themselves, in order to avoid social sanctions and isolation. The subsequent overrepresentation of the alleged majority leads to misperception of the opinion climate, which reinforces the silence of opinion groups that seem to be the minority. This pattern has been generally supported (Glynn, Hayes, & Shanahan, 1997) and also aligns with recent evidence that exposure to attitude-consistent information fosters political engagement and activity (Dilliplane, 2011; Stroud, 2007, 2010), while exposure to counterattitudinal information, whether mediated or interpersonal, depresses participation (Dilliplane, 2011; Matthes, 2012; McClurg, 2006; Mutz, 2002; Nir & Druckman, 2008). Spiral of silence theorizing also touched upon a topic that has garnered much recent research interest in the Internet use context selective exposure. Commenting on individuals who are not swayed in their opinion despite a perceived disagreeable majority, Noelle-Neumann (1974, p. 49) noted many of them may manage to support their opinions by selecting out persons and media which confirm their views. This rationale can be extended to all opinion holders, so that those who practice Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 185

more attitude-consistent selectivity will, as a consequence, perceive their attitudes as especially better represented in public opinion, while those who exhibit less confirmation bias in their selective exposure will encounter a wider variety of counterattitudinal opinions. The former scenario would thus foster polarization (Stroud, 2010) and less consideration of other opinions, whereas the latter scenario features what Lazarsfeld, Berelson, and Gaudet (1944) characterized as cross-pressures that foster silence and restraint from political engagement. Selective Exposure in the Internet Context Selective exposure was initially described by Lazarsfeld et al. (1944) and given theoretical grounding by Festinger s cognitive dissonance theory (1957). With regard to democratic outcomes, individuals tendency to isolate themselves by selecting only attitude-consistent news is feared to have detrimental effects for political values like knowledge, moderation, and deliberation (Kull, Ramsay, & Lewis, 2003; Sunstein, 2001). With the Internet as a new and important context in which selective exposure might occur, a recent debate has ensued among communication scholars. While Bennett and Iyengar (2008, 2010) have suggested that, in the Internet age, media users are moving into a new era of limited effects in which selectivity reduces the likelihood of media-driven persuasion or shifts in opinion (Klapper, 1960; Noelle-Neumann, 1974), Holbert et al. (2010) argued that selectivity may produce substantial media effects. Some empirical findings demonstrate selective exposure to attitude-consistent online messages, but little avoidance of counterattitudinal online messages (Garrett, 2009a, 2009b; Johnson, Zhang, & Bichard, 2011; Knobloch-Westerwick & Kleinman, 2012). Moreover, heavy Internet users show less loyalty to specific news sites (Purcell, Rainie, Mitchell, Rosentiel, & Olmstead, 2010). The sheer volume of political news that more habitual online news users are exposed to may mean that more cross-pressure exposure is inevitable and that they are accustomed to heterogeneous views in their media use. This reasoning leads to our first hypothesis. H1: More habitual online news use decreases attitude-consistent selective exposure. This hypothesized relationship is essential, as selective exposure should mediate the relationship between online news habit and political participation. If more counterattitudinal exposure occurs online, political participation is likely to be suppressed as a result of online information exposure, as much empirical evidence suggests: Contact with cross-pressures has been found to reduce the likelihood of voting (Mutz, 2002), delay voting decisions (Dilliplane, 2011; Matthes, 2012; McClurg, 2006; Mutz, 2002; Nir & Druckman, 2008), and depress political discussion, campaigning, donating, and meeting attendance (Giles & Dantico, 1982; Huckfeldt, 1979; McClurg, 2006; Mutz, 2002). In contrast, effects of confirmation-biased selective exposure include increases in attitude extremity (Stroud, 2010), accessibility of political self-concept (Knobloch-Westerwick & Meng, 2011), campaign participation (Dilliplane, 2011), and political discussion (Stroud, 2007), along with decreased time for voting decision (Dilliplane, 2011). This empirical evidence, along with spiral of silence theorizing, leads to the second hypothesis. H2: Greater attitude-consistent selective exposure increases political participation. Building on H1 and H2, the third hypothesis suggests a mediating role of selective exposure for the impact of online news use habit on political participation. 186 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

H3: Attitude-consistent selective exposure mediates the effect of habitual online news use on political participation. Political Interest When examining possible impacts of online news use on selective exposure and political participation, it is important to consider potential confounding variables. A likely suspect is political interest studies of media s influence on political participation have documented the relationship between political interest and subsequent news media use, with either mass media (McLeod et al., 1996; McLeod, Scheufele, & Moy, 1999) or online news (see Boulianne, 2009) acting as a mediator between interest and engagement. However, as Lazarsfeld et al. (1944) observed, and more recent studies of Internet use continue to find, a reciprocal relationship exists between interest and media use. Thus consuming news media may increase political interest even further (Boulianne, 2009, 2011). Yet political interest can be thought to precede political participation and is modeled accordingly in the present research. In the full model, the effect of political interest on participation will be sequentially mediated by online news use habit and attitude-consistent selective exposure. H4: Greater political interest increases political participation. H5: Greater political interest increases habitual online news use. H6: Habitual online news use and attitude-consistent selective exposure sequentially mediate the effect of political interest on political participation. Method Overview A two-session online field study was administered to 205 nonstudent, U.S. adults in the month prior to the 2008 presidential election. The initial session captured measurements for political interest, new use habit, and attitudes toward 12 political issues. These same participants returned to the study website for a second session several days later. They then browsed an online news magazine, featuring eight stories on four topics, with both an attitude-consistent and counterattitudinal story for each topic. The topics were among the 12 political issues for which attitudes had already been measured. The experimental site software unobtrusively logged exposure to each story, after which time participants responded to survey items measuring likelihood of political participation concerning these story topics. Respondents The sample consisted of 205 nonstudent adults. About half of participants (53%) were female, and the average age was 40.57 (SD = 10.17), with no gender difference in age, t(202) = 0.314, p =.75. Participants were 85% White, 7% Black, 5% Asian, and 3% Hispanic. Median household income was $50 60K, while highest level of education was 37% high school diploma, 41% bachelor s degree, and 22% graduate degree. Most (70%) were employed, with 9% self-employed, 10% homemakers, and 11% either retired or unable to work. Seventy-two percent were married, 9% were divorced, 17% never married, with the remainder dating, separated, or widowed. The mailing addresses for incentives suggested that 42 individuals (21% of participants) were married couples. Less than half (44.88%) were from a single Midwestern state, while others resided across 27 different U.S. states. Six participants were removed from analysis because they spent 3.5 minutes or more on the overview page of the online Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 187

magazine, suggesting they were not engaged in browsing the site. This produced a final sample size of 199 for hypothesis testing. Procedure Data from the two sessions were gathered online from September 28 to November 3, 2008, just prior to the U.S. general election. In an unconventional recruitment approach, both social and financial incentives were utilized to motivate participants, in order to obtain a diverse sample regarding economic, educational, and psychological characteristics. A two-step process was used to recruit the sample, in which students from a large Midwestern U.S. university received extra course credit for recruiting nonstudent participants aged 30 65. Students were instructed to enroll up to five individuals, and they averaged M = 2.22 (SD = 0.91) successful recruits. The actual adult participants received a $20 check for completing the two-session study. Students were contacted by their instructors about recruiting non-students aged 30 65 for a study regarding news outlets and judgments of news issues. Students were provided with a website that facilitated recruitment of study participants by e-mail. The prospective participants were told about the study of news outlets and judgments of news issues, and offered a $20 incentive for their involvement. They were then directed to install a browser plug-in that would allow the research application to execute. The research site was programmed by the first author using Authorware7, a software-authoring tool (for details, see Adobe, n.d.; Tew & McGraw, 2002). Participants followed a web link to begin the first session on the research site. The landing page notified participants that full attention was required, and that they could quit and take the session later to avoid distractions. The first session solicited issue attitudes, political interest, news use habits, and demographics. The second session took place four days after the first. Participants each received an e-mail with a link to the second session, and were once again reminded that full attention was needed during the session. Once participants accessed the session and agreed to commit their attention to the task, an online news magazine was presented for four minutes, during which time the participants were free to the browse the eight articles. Four minutes is typical of the amount of time online readers spend with online newspaper sites: on average about 3.6 minutes per visit, and 50 seconds per page (Newspaper Association of America, n.d.). Therefore, study participants typically should have been able to view around half of the story pages, ensuring that selectivity would occur and that the browsing task resembled usual online news use. After the 4 minutes of browsing had elapsed, participants were directed to a questionnaire. They were asked to rate articles, on 7-point scales from not at all to extremely, on how credible/important/biased/interesting/well-written/relevant they were. These questions merely served to provide closure for the browsing task and corresponded to the ostensible purpose of the study. Attitudes towards the target political issues were measured again, followed by questions about the likelihood that the respondent would be willing to engage in political participation concerning these issues. Participants were then debriefed, thanked, and provided with details regarding receipt of their incentives. Stimulus Online Magazine Display of available articles. The online news magazine was designed to resemble popular online news magazines. A masthead with the name and logo America s national forum: Online opinion appeared at the top of the site. A deactivated navigation bar on the left side of the page listed section headings like Economics and Science, as commonly found on online news magazines. The home page listed previews for all articles, arranged into two columns. 188 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

On the home page, each story preview consisted of a headline, news lead, and hyperlink to the full story. Each lead was approximately the same length, with a range of 24 28 words. Story previews were randomized in their placement, to prevent any position effects. The randomization procedure was, however, constrained to prevent the adjacent placement of two opposing articles on the same topic. Respondents could choose stories by clicking on their hyperlinks, which led to an individual page for each article. They would then return to the home page to choose other articles. Each time a participant accessed or left an article page, their hyperlink click was logged, providing data on time spent with each page. News leads and article texts. The four issues selected gun control, abortion, minimum wage, and health care were relevant issues in the 2008 election, and the pretest demonstrated that they elicited strong attitudes in opposing directions. Stimuli texts were adapted from partisan and advocacy group websites, with two opposing articles and headlines for each topic, which were pretested to be equally interesting. The topic headlines were (a) Universal health care and Personalized health coverage, (b) Firearm threat and Self-defense rights, (c) Cruelty of pro-choice and Abortion is pro-life, and (d) Increase minimum wage and Wage raising hurts. Articles were edited to achieve equivalent length, 705 719 words (M = 716, SD = 5.2), and Flesch reading ease scores were similar (36 and 36 for minimum wage, 33 and 37 for health care, 53 and 47 for gun control, 56 and 51 for abortion, on a 0 100 scale). All of the articles utilized editorial rhetoric and details like statistics, dates, budget figures, and descriptions of policies, but none of the articles used exemplars. Commonly cited sources in articles were researchers (e.g., institutes or scholarly journals) and generally venerable and neutral agencies like the Census Bureau, World Health Organization, or FBI. Some articles referred to advocacy groups like the Center to Prevent Handgun Violence or the union movement America Needs a Raise. References to political parties were minimal, and usually balanced. Measures Attitudes. In both sessions, dichotomous measures were used to capture attitudes towards political issues and allowed for assignment of news stories as either attitude-consistent or counterattitudinal. Four target issues (health care, gun control, abortion, and minimum wage) were presented, randomly mixed in with eight filler issues. Participants responded by choosing oppose or support to short policy statements. Table 1 provides for the distribution of dichotomous attitude measures across the sample. Attitude measures were reliable across sessions, and were validated with Likert-scale measures Table 1 Descriptive Statistics of Dichotomous and Likert-Scale Attitude Measures Issue Attitude % support Session 1 Session 2 Attitude (1=support, 5=oppose) M(SD) Attitude % support Attitude (1=support, 6=oppose) M(SD) Stricter Gun Control 67 2.29 (1.5) 65 2.93 (1.9) Legal Abortion Pro-Choice 57 2.76 (1.74) 62 3.23 (2.1) Universal Health Care 69 2.31 (1.6) 68 2.78 (1.9) Increase Minimum Wage 86 1.79 (1.2) 83 2.32 (1.5) Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 189

similar to those used by the American National Election Studies (ANES, n.d.). Likert-scale measures were correlated with dichotomous measures for both sessions, r =.71 and.82 (all p <.001). Political interest. Toward the end of the first session, participants were presented with a question adopted from the ANES studies (ANES, n.d.), How closely do you follow what s going on in government and public affairs? with a 4-point scale and the response options not closely at all/somewhat closely/not too closely/very closely. The average score was 3.17 (SD = 0.69). Online news use habit. At the end of the first session, participants rated the frequency of their use of online news, daily newspapers, TV news, political websites, and talk/comedy shows about news and politics on a 6-point scale, every day/several times a week/once a week/several times a month/once a month/less often, coded with higher values for greater frequency. Frequency of online news had an average of 3.39 (SD = 1.72). Selective exposure measures. During the browsing task in the second session, the software logged each hyperlink to track participant browsing, so that exposure was measured both as article selection (whether a given article was read or not) and reading time in seconds. Most importantly, session 1 dichotomous attitude measurements allowed each article to be categorized as attitude-consistent or counterattitudinal exposure at the individual level. Participants spent an average of 71.64 seconds (SD = 55.87) on the overview page. They selected an average of 3.85 articles (SD = 2.11), with an average of 47.95 seconds (SD = 56.08) spent on the topic of abortion, 33.77 seconds (SD = 46.24) on gun control, 36.34 seconds (SD = 43.63) on the minimum wage, and 48.80 seconds (SD = 53.28) on health insurance. With regard to exposure as it related to attitudes, more attitude-consistent articles were chosen (M = 2.10, SD = 1.20) than counterattitudinal articles (M = 1.81, SD = 1.29), t(198) = 3.17, p =.002, and exposure in seconds was greater for attitude-consistent articles (M = 94.33, SD = 61.77) than counterattitudinal articles (M = 72.54, SD =56.39), t(198) = 2.95, p =.004, across the four topics. Given that some research has found that selective exposure to consonant information is a stronger tendency than avoidance of dissonant information (Frey, 1982; Garrett, 2009b), selective exposure was split into two indicators, one that averaged selection of attitude-consistent articles and minutes spent on those articles (M = 1.89, SD = 0.80; the correlation between the selection and exposure time was r =.09, p =.197) and one that averaged selection of counterattitudinal articles and minutes spent on those articles (M = 1.54, SD = 0.89; the correlation between selection and exposure time was r =.34, p <.001). The attitude-consistent and counterattitudinal indices were negatively correlated, r =.33 (p <.001). Political participation likelihood. At the end of the second session, participants were asked about their willingness to engage in several types of political participation for the target issues. For each issue, study participants were asked: (a) Would you sign a petition in favor of the <ISSUE> movement? (b) Would you go to meetings, speeches, fund raising events, or things like that in support of the <ISSUE> movement? (c) Would you wear a button or put a sticker on your car or place a sign in your window or in front of your house to support the <ISSUE> movement? (d) Would you say yes if you were asked to help staff an information table on campus to support the <ISSUE> movement? (e) Would you vote in favor of a ballot issue about the <ISSUE>? These were customized for each participant to phrase the issue with regard to his or her stance on the issue. Issues were characterized as minimum wage increase or minimum wage freeze, gun control or gun rights, universal healthcare or personalized health care, and prochoice or prolife, based on the participant s dichotomous attitude measure in session 1. Willingness to participate in each of the five activities was measured with a 6-point scale anchored by 0 = certainly not and 5 = most certainly. An index of political participation likelihood was 190 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

calculated by summing the scores across all 4 topics and all 5 types of participation, generating a scale ranging from 0 100 (M = 63.35, SD = 20.65; Cronbach s alpha =.92). Results The hypotheses were tested simultaneously with a path model in MPLUS 5.1, using bootstrapping to test indirect effects (see Preacher & Hayes, 2008). The results for nonstandardized coefficients and indirect effects for each path are illustrated in Figure 1. In the model, more habitual online news use decreased selective exposure to attitude-consistent information (b = 5.94, SE = 2.42, p =.014), supporting H1. Selective exposure to attitude-consistent information increased the likelihood of political participation (b = 0.08, SE = 0.02, p <.001), corroborating H2. The impacts of the negative effect in H1 and positive effect in H2 worked together to reduce the likelihood of political participation, with selective exposure mediating the relationship between online news use habit and participation. Thus H3 was supported with a significant indirect effect (b = 0.47, SE = 0.23, 95% CI [ 1.08, 0.11]). There was no significant direct effect of online news use habit on participation (b = 0.97, SE = 0.89, p =.27). With regard to H4 and H5, greater political interest led directly to greater likelihood of political participation (b = 6.34, SE = 2.08, p =.002) and to more frequent online news use (b = 0.48, SE = 0.17, Figure 1 Direct and indirect effects on political participation likelihood. Confidence intervals created with 5000 bootstrap resamples. Standard error in parentheses. *p <.05. **p <.01. ***p <.001. Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 191

p =.005), supporting both hypotheses. The path from interest to selective exposure (b = 1.51, SE = 6.57, p =.82) was not significant. It was evident that selective exposure mediated the effect of online news use habit to depress political participation, supporting H3. It was also predicted that in addition to interest s direct effect on participation, online news use habit and selective exposure would act through sequential mediation to affect participation. The specific indirect mediation through all four variables (interest to news habit to selective exposure to participation) was significant (b = 0.23, SE = 0.14, 95% CI [ 0.67, 0.05]), supporting H6. The mediators worked to reduce political participation from a total effect of b = 6.70, SE = 2.07, p =.001 to a direct effect of b = 6.34. Finally, a model using the index of counterattitudinal exposure failed to generate significant effects. While increased use of online news decreased selective exposure, it did not influence exposure to dissonant information in a way that would mediate participation effects, too. Discussion The present study examined the role that selective exposure played in the effect of online news use on political participation. The expected relationships were supported by the mediation analysis. The findings demonstrate that habitual online news use decreased the likelihood of engaging in attitudeconsistent selective exposure (H1), selective exposure increased political participation likelihood (H2), and that selective exposure acted a mediator, decreasing participation as online news use became more frequent (H3). Political interest directly increased participation (H4) and increased online news use (H5). Online news use and selective exposure sequentially mediated the effect of political interest, working to diminish political participation (H6). As the direct effect of political interest was positive, this mediation pattern is an instance of inconsistent mediation (MacKinnon, Krull, & Lockwood, 2000), in which the indirect effect works against the direct effect. In other words, even though political interest fostered online news use and political participation, greater habitual online news reduced selective exposure to attitude-consistent messages, which in turn undermined political participation. Thus, even though it is often thought to be desirable that citizens in a democracy attend to diverse and cross-cutting messages that do not align with their preexisting views, such exposure also hinders participation. Hence, selective exposure driven by a confirmation bias may affect the democratic process for better or worse if it is present, less consideration of diverse viewpoints occurs, even though such consideration is at the heart of the democratic process. Its presence, however, advances political participation, which is essential to accomplish plurality in opinions. These findings offer important evidence regarding the effects of selective exposure on political participation and of online news use habit on selective exposure, along with critical implications for both the study of the influence of the Internet on political engagement and the ongoing debate over selective exposure and media effects. Each will be considered in turn, followed by a consideration of the study s limitations and suggestions for future directions. In keeping with previous findings (Dilliplane, 2011; Stroud, 2007, 2010), participation increased when participants engaged in selective exposure to attitude-consistent information. Reduced selective exposure to consonant information, which was driven by habitual online news use, diminished participation. Given recent findings that there is little avoidance of dissonant online information (Garrett, 2009a, 2009b; Johnson et al., 2011; Knobloch-Westerwick & Kleinman, 2012), we might expect that those who consume large amounts of online news and do not exhibit strong selectivity are experiencing cross-pressures in their news environment, which reduces their likelihood of participating in the political system. 192 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

By incorporating selective exposure into the model of Internet use and political engagement, we are able to offer an explanation for the generally weak and insignificant effect on political participation (Boulianne, 2009), despite common expectations for a stronger positive effect. As online news use increases, selectivity is decreasing, diminishing participation. Selectivity is an important step in, as Lazarsfeld et al. (1944) termed it, activating political behavior. It allows the individual to maintain or enhance the perception that their opinion group is dominant and that they can freely engage in opinion expression (Noelle-Neumann, 1974). Yet frequent online news use reduces the likelihood that selective exposure, and subsequent expression, will occur. The literature on media use and political engagement assumes that the breadth and depth of online information will accommodate more diverse deliberation and discussion. This greater diversity may indeed emerge, but spiral of silence theory and findings regarding cross-pressures and network heterogeneity make it clear that the resulting cross-pressure reduce desirable engagement outcomes (Giles & Dantico, 1982; Matthes, 2012; McClurg, 2006; Mutz, 2002; Nir & Druckman, 2008). Increased exposure to divergent views through Internet use, good from a normative perspective, creates a dilemma in which participation is suppressed, whether through ambivalence (Mutz, 2002), social pressure, or the lack of reinforcing activation (Lazarsfeld et al., 1944). Our findings also have important implications for recent debate over whether fragmentation and polarization caused and facilitated by selective exposure are leading us into a new era of minimal effects (Bennett & Iyengar, 2008, 2010), or whether selectivity presents new opportunities for significant and meaningful media effects (Holbert et al., 2010). Our findings document a limited effects situation (i.e., diminished political engagement) not because of selective exposure, but rather because of the absence of a selective bias in media choice for habitual users of online news. It appears that although attitude-consistent selective exposure does have the potential to drive desirable effects, people are less likely to exercise this selectivity when they are habitual online news users. This study identified an inconsistent mediation effect by which selective exposure influences the Internet use and political engagement relationship, and it provided evidence for the relationships between online news use and selectivity as well as between selectivity and political participation. The sequential measures allowed for some level of control and causal inference; the unobtrusive measure of selective exposure avoids problems associated with self-reports of media use. However, limitations must be noted. First, as a non-experimental design, the ability to make causal claims is necessarily restricted. Selective exposure is a naturally occurring behavior, as is the extent of online news habit, and cannot be assigned in a true experiment. However, by giving participants the opportunity to engage in this behavior in this lab, capturing behavior in temporal sequence, and demonstrating the theoretical rationale behind the casual order, we are able to infer causal relationships with some certainty. Second, the sample, while an adult sample exhibiting a fairly representative set of demographic variables, is neither truly representative nor randomly selected. The two-step recruitment process could have possibly introduced bias into the sample composition in a way that would influence the findings. Finally, the study assessed selective exposure and participation with regard to political issues. The findings may not necessarily generalize to political parties and candidates in the same way. Future research should consider selective exposure as a mediator between online news use and political engagement. Both experimental and survey data are needed to replicate and shed further light on this finding, and a range of political engagement measures should be utilized, from discussion to voting. It would also be useful to test whether the process holds regardless of whether the political engagement is issue-oriented, as in this study, or party or candidate-oriented. More experimental and panel survey work is needed to fully understand the causal relationship between political interest and news use, and perhaps even between selective exposure and participation, and any implications that reciprocal relationships might have for theory. Clearly, theoretical development is needed in this area. Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 193

Researchers examining media use and engagement have not paid heed to the activating role of selective exposure that Lazarsfeld et al. identified 70 years ago, and those who have examined the negative effects of cross-pressures have been hesitant to examine the positive effects of attitude-consistent exposure. Spiral of silence theory provides a larger framework for understanding these effects as part of the same process. Perhaps the perception that spiral of silence only explains literal speaking out has discouraged its use as a larger framework (see Glynn et al., 1997). Finally, an important direction for future research is to understand why and how online news use works to reduce selective exposure behavior. In conclusion, selective exposure is a crucial step between online news use and political engagement. As online news use habit increases, the tendency towards selective exposure to attitude-consistent material declines, which subsequently reduces participation. Political participation is a desirable outcome, which is influenced by exercising selectivity and consuming attitude-consistent messages. An online news use habit, driven by the same independent variable of political interest, presents a dilemma, as it will impair the likelihood by which an individual will get involved in the political system and be an engaged citizen. References Adobe. (n.d.). Authorware. Retrieved from http://www.adobe.com/products/authorware/ ANES. (n.d.) Questions asked in ANES surveys [WWW document].retrieved from http://www. electionstudies.org/resources/questions/questions.htm Benkler, Y. (2006). The wealth of networks: How social production transforms markets and freedom.new Haven, CT: Yale University Press. Bennett, W. L., & Iyengar, S. (2008). A new era of minimal effects? The changing foundations of political communication. Journal of Communication, 58(4), 707 731. doi: 10.1111/j.1460-2466.2008.00410.x Bennett, W. L., & Iyengar, S. (2010). The shifting foundations of political communication: Responding to a defense of the media effects paradigm. Journal of Communication, 60(1), 35 39. doi: 10.1111/ j.1460-2466.2009.01471.x Bimber, B. (1998). The Internet and political transformation: Populism, community, and accelerated pluralism. Polity, 31(1), 133 160. doi: 10.2307/3235370 Boulianne, S. (2009). Does Internet use affect engagement? A meta-analysis of research. Political Communication, 26(2), 193 211. doi: 10.1080/10584600902854363 Boulianne, S. (2011). Stimulating or reinforcing political interest: Using panel data to examine reciprocal effects between news media and political interest. Political Communication, 28(2), 147 162. doi: 10.1080/10584609.2010.540305 Dilliplane, S. (2011). All the news you want to hear: The impact of partisan news exposure on political participation. Public Opinion Quarterly, 75(2), 287 316. doi: 10.1093/poq/nfr006 Festinger,L. (1957).A theory of cognitive dissonance. Stanford, CA: Stanford University Press. Frey, D. (1982). Different levels of cognitive dissonance, information seeking, and information avoidance. Journal of Personality and Social Psychology, 43(6), 1175 1183. doi: 10.1037/0022-3514. 43.6.1175 Garrett, R. K. (2009a). Echo chambers online? Politically motivated selective exposure among Internet news users. Journal of Computer-Mediated Communication, 14(2),265 285.doi:10.1111/j.1083-6101.2009.01440.x Garrett, R. K. (2009b). Politically motivated reinforcement seeking: Reframing the selective exposure debate. Journal of Communication, 59(4), 676 699. doi: 10.1111/j.1460-2466.2009.01452.x 194 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association

Giles, M. W., & Dantico, M. K. (1982). Political participation and neighborhood social context revisited. American Journal of Political Science, 26(1), 144 150. doi: 10.2307/2110844 Glynn, C. J., Hayes, A. F., & Shanahan, J. (1997). Perceived support for one s opinions and willingness to speak out: A meta-analysis of survey studies on the spiral of silence. Public Opinion Quarterly, 61(3), 452 463. doi: 10.1086/297808 Holbert, R. L., Garrett, R. K., & Gleason, L. S. (2010). A new era of minimal effects? A response to Bennett and Iyengar. Journal of Communication, 60(1), 15 34. doi: 10.1111/j.1460-2466. 2009.01470.x Huckfeldt, R. R. (1979). Political participation and the neighborhood social context. American Journal of Political Science, 23(3), 579 592. Johnson, T. J., Zhang, W., & Bichard, S. L. (2011). Voices of convergence or conflict? A path analysis investigation of selective exposure to political websites. Social Science Computer Review, 29(4), 449 469. doi: 10.1177/0894439310379962 Klapper, J. T. (1960). The effects of mass communication. New York, NY: The Free Press. Knobloch-Westerwick, S., & Kleinman, S. B. (2012). Preelection selective exposure: confirmation bias versus informational utility. Communication Research, 39(2), 170 193. doi: 10.1177/00936502 11400597 Knobloch-Westerwick, S., & Meng, J. (2011). Reinforcement of the political self through selective exposure to political messages. Journal of Communication, 61(2), 349 368. doi: 10.1111/ j.1460-2466.2011.01543.x Krueger, B. S. (2002). Assessing the potential of Internet political participation in the United States: A resource approach. American Politics Research, 30(5), 476 498. doi: 10.1177/1532673X 02030005002 Kull, S., Ramsay, C., & Lewis, E. (2003). Misperceptions, the media, and the Iraq War. Political Science Quarterly, 118(4), 569 598. Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1944). The people s choice: How the voter makes up his mind in a presidential campaign. New York: Columbia University Press. MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the mediation, confounding, and suppression effect. Prevention Science, 1(4), 173 181. doi: 10.1023/A:1026595011371 Matthes, J. (2012). Exposure to counterattitudinal news coverage and the timing of voting decisions. Communication Research, 39(2), 147 169. doi: 10.1177/0093650211402322 McClurg, S. D. (2006). The electoral relevance of political talk: Examining disagreement and expertise effects in social networks on political participation. American Journal of Political Science, 50(3), 737 754. doi: 10.1111/j.1540-5907.2006.00213.x McLeod, J. M., Daily, K., Guo, Z., Eveland, W. P., Jr., Bayer, J., Yang, S., & Wang, H. (1996). Community integration, local media use, and democratic processes. Communication Research, 23(2), 179 209. doi: 10.1177/009365096023002002 McLeod, J. M., Scheufele, D. A., & Moy, P. (1999). Community, communication, and participation: The role of mass media and interpersonal discussion in local political participation. Political Communication, 16(3), 315 336. doi: 10.1080/105846099198659 Mutz, D. C. (2002). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46(4), 838 855. doi: 10.2307/3088437 Newspaper Association of America. (n.d.). Aggregated U.S. newspaper web site audience. Retrieved from http://www.naa.org/trends-and-numbers/newspaper-websites/newspaper-web-audience.aspx Nir, L., & Druckman, J. N. (2008). Campaign mixed-message flows and timing of vote decision. International Journal of Public Opinion Research, 20(3), 326 346. doi: 10.1093/ijpor/edn035 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association 195

Noelle-Neumann, E. (1974). The spiral of silence: A theory of public opinion. Journal of Communication, 24(2), 43 51. doi: 10.1111/j.1460-2466.1974.tb00367.x Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879 891. doi: 10.3758/BRM.40.3.879 Purcell, K., Rainie, L., Mitchell, A., Rosenstiel, T., & Olmstead, K. (2010). Understanding the participatory news consumer: How Internet and cell phone users have turned news into a social experience. Washington, DC: The Pew Internet & American Life Project. Stroud, N. J. (2007). Media effects, selective exposure, and Fahrenheit 9/11. Political Communication, 24(4), 415 432. doi: 10.1080/10584600701641565 Stroud, N. J. (2010). Polarization and partisan selective exposure. Journal of Communication, 60(3), 556 576. doi: 10.1111/j.1460-2466.2010.01497.x Sunstein, C. R. (2001). Republic.com. Princeton, NJ: Princeton University Press. Taylor, A. B., MacKinnon, D. P., & Tein, J.-Y. (2008). Test of the three-path mediated effect. Organizational Research Methods, 11(2), 241 269. doi: 10.1177/1094428107300344 Tew, M. D., & McGraw, K. O. (2002). The accuracy of response timing by Authorware programs. Retrieved from http://www.psych.uni.edu/psychexps/scrapbook/timing_2002.pdf Weber, L. M., Loumakis, A., & Bergman, J. (2003). Who participates and why? An analysis of citizens on the Internet and the mass public. Social Science Computer Review, 21(1), 26 42. doi: 10.1177/0894439302238969 About the Authors Silvia Knobloch-Westerwick (Ph.D., University of Music, Drama & Media Hanover, Germany) is an Associate Professor at the School of Communication, The Ohio State University. Her research interests include media effects and selective exposure to media messages in the contexts of news, health information, entertainment, political communication, and new communication technologies. Address: 3020 Derby Hall, 154 N. Oval Mall, Columbus, OH 43210 Benjamin K. Johnson (MA, Michigan State University) is a Doctoral Student and Graduate Associate at the School of Communication, The Ohio State University. His research is focused on selective exposure in new media settings, especially as it relates to social comparison and impression management. Address: 3016 Derby Hall, 154 N. Oval Mall, Columbus, OH 43210 196 Journal of Computer-Mediated Communication 19 (2014) 184 196 2013 International Communication Association