Abstract
Selective exposure research indicates that news consumers tend to seek out attitude-consistent information and avoid attitude-challenging information. This study examines online news credibility and cognitive dissonance as theoretical explanations for partisan selective exposure behavior. After viewing an attitudinally consistent, challenging, or politically balanced online news source, cognitive dissonance, credibility perceptions, and likelihood of selective exposure were measured. Results showed that people judge attitude-consistent and neutral news sources as more credible than attitude-challenging news sources, and although people experience slightly more cognitive dissonance when exposed to attitude-challenging news sources, overall dissonance levels were quite low. These results refute the cognitive dissonance explanation for selective exposure and suggest a new explanation that is based on credibility perceptions rather than psychological discomfort with attitude-challenging information.
In a speech critiquing the news media, President Barack Obama said, “Today’s 24/7 echo chamber amplifies the most inflammatory sound-bytes louder and faster than ever before” (Johnson, 2010). He went on to implore the U.S. public to seek out a balanced news diet, saying that “if we choose only to expose ourselves to opinions and viewpoints that are in line with our own, we become more polarized . . . That will only reinforce and even deepen the political divides in our country” (Johnson, 2010). Scholars have labeled this behavior on the part of news audiences as partisan selective exposure, which is the tendency for people to seek out news information or sources that share their own political viewpoints (Lazarsfeld, Berelson, & Gaudet, 1944).
Research on partisan selective exposure began decades ago, but recent changes in the news media environment have reignited interest in this phenomenon. Changes include a proliferation of news sources available to the public from relatively few to comparatively many, the structure of news providers from centralized and monolithic to decentralized and diversified (although of course many news outlets are still controlled by huge conglomerates), and the ability of media audiences to control their own exposure to news from among the many sources that are now available via digital media (Metzger & Chaffee, 2001). To stand out from the cacophony, many news sources began to target niche groups in order to draw an audience (Nelson-Field & Riebe, 2011). Stroud (2011), for example, provided evidence that U.S. news media are becoming more biased in order to appeal to partisan audiences.
Although the proliferation of politically biased news sources is alarming to some, their existence and popularity do not by themselves provide evidence of a selective exposure effect. People can still get a balanced news diet by using nonpartisan news sources or by visiting multiple news sources to support multiple political ideologies, if they choose to do so. However, studies suggest that news consumers are not taking advantage of the diverse resources at their fingertips. Research on partisan selective exposure in the modern news era has found more consistent evidence of biased news consumption than did research in the pre-Internet media era. For example, scholars have found that nearly a quarter of news consumers almost exclusively use news sources that share their point of view (Kohut, Doherty, Dimock, & Keeter, 2012), many rely on at least one attitude-consistent news source (Stroud, 2008, 2011), and people are more likely to view a news item if the headline suggests attitude-confirming information (Garrett, 2009a; Knobloch-Westerwick & Meng, 2009; Westerwick, Kleinman, & Knobloch-Westerwick, 2013).
Despite the evidence for partisan selective exposure, questions remain about the causal mechanisms that underlie the phenomenon. Many studies have employed cognitive dissonance theory as a causal explanation. Festinger (1957) postulated that people desire consistency among their attitudes as well as between their attitudes and their behaviors. When people are aware of inconsistencies, they experience cognitive dissonance, which causes mental discomfort that they are motivated to avoid and reduce. Although many researchers argue that news consumers turn to like-minded information sources as a means to avoid or reduce cognitive dissonance (Brannon, Tagler, & Eagly, 2007; Garrett, 2009a, 2009b; Iyengar & Hahn, 2009; Jonas, Schulz-Hardt, Frey, & Thelen, 2001; Knobloch-Westerwick & Meng, 2009), no studies actually measure whether news consumers indeed experience cognitive dissonance when confronted with attitude-challenging information. As a result, it is impossible to tell from past research whether dissonance avoidance or reduction is what motivates partisan selective exposure, or whether there is another explanation for the phenomenon. 1
An alternative explanation for partisan selective exposure comes from news consumers’ evaluations of the credibility of a news source and/or the information within news stories. Rather than seeking to reduce dissonance, it may be that individuals seek information from news sources that share their political attitudes and outlook (i.e., attitude-consistent news sources) because they perceive them as more credible than either sources that report from a political perspective that differs from that of the audience member (i.e., attitude-challenging news sources) or from news sources that provide information from a variety of political perspectives. Although some research on source credibility finds that people judge unbiased news sources as more credible than biased or one-sided news sources (Allen, 1991; Pechmann, 1992; Zhao & Capella, 2008), other research suggests that people may judge biased but attitudinally consistent sources as more credible than unbiased sources (Clark & Maass, 1988; Mackie & Queller, 2000; Vallone, Ross, & Lepper, 1985).
Indeed, credibility could provide an entirely new explanation for partisan selective exposure, and so it is surprising that in spite of its relevance to information selection and attention, little research has been conducted addressing the role of credibility perceptions in selective exposure (but see Johnson & Kaye, 2013; Knobloch-Westerwick, Callison, Chen, Fritzsche, & Zillman, 2005). Thus, the goal of this study is to examine how credibility relates to selective exposure in the partisan media environment. Specifically, this study will look at how individuals evaluate the credibility of biased news sources and stories, and the role of both cognitive dissonance and credibility perceptions in selective exposure to attitude-consistent news information.
Selective Exposure to Attitude-Consistent Information
Several early reviews of the selective exposure literature (Cotton, 1985; Sears & Freedman, 1967) concluded that there was not sufficient evidence for the phenomenon: While some studies indicated that individuals have a preference for attitude-consistent information, others found that individuals neither seek out like-minded sources nor avoid attitude-challenging sources. However, these early studies took place in a media environment characterized by comparatively fewer information sources and more balanced news reporting than exists today (Stroud, 2011). More recently, researchers have found consistent support of partisan selective exposure by news consumers.
In a survey of a representative sample of U.S. news consumers, Stroud (2008) found that 64% of Republicans consistently relied on at least one conservative news source, and 26% of Democrats consistently used a conservative source (see also Stroud, 2011). Moreover, 76% of liberals relied on at least one liberal source, compared with 43% of conservatives. Iyengar and Hahn (2009) found that, when given a choice among five news sources (Fox News, CNN, BBC, NPR, and an unattributed source), conservative participants significantly preferred Fox News over any other source, while liberal participants avoided Fox News. More recently, a study by the Pew Research Center (2014) similarly found a strong preference among conservatives for Fox News, and a preference among liberals to use NPR, MSNBC, and the New York Times as their top source for political news.
The selective exposure phenomenon may be even more pronounced online because consumers have greater control over which sources and messages they are exposed to and because, for many online political media sources, maintaining readership seems contingent on taking a side. Accordingly, there has been empirical support for the existence of partisan selective exposure in the blog context. Johnson, Bichard, and Zhang (2009) found that blog users have a tendency to visit blogs that share their political predispositions and avoid blogs that challenge them, as predicted by the selective exposure hypothesis. And, these effects are not limited to blogs, as evidence for selective exposure to confirmatory political messages has emerged across various online media in recent years (Knobloch-Westerwick, 2014).
Research on selective exposure also indicates that selection processes operate at both the source and message levels in online news generally. Political bias at both the source level (i.e., political bias of a news website or blog) and story level (i.e., whether a news story presents one political perspective exclusively or more strongly) appear to impact people’s decision when selecting a news source to view, as well as how long they spend viewing the story information online. Garrett (2009a) found that when given a choice among attitude-consistent and attitude-challenging news stories, individuals were more likely to view news stories if they thought they would confirm their opinions, and experienced a slight aversion to selecting information that appeared to disconfirm their opinions. Similarly, Knobloch-Westerwick and Meng (2009) found that people chose attitude-consistent sources of news information online significantly more often than counter-attitudinal ones when cued to story bias by article headlines, and spent more time reading attitude-consistent stories after choosing them. In this study, individuals were more likely to seek out attitude-consistent information from online sources that they used frequently and if they felt strongly about their beliefs (see also Brannon et al., 2007; Kobayashi & Ikeda, 2009; Westerwick et al., 2013).
However, while recent studies certainly provide support for the selective exposure phenomenon, there has been little advancement in the application of theory to selective exposure research. A direct test of selective exposure’s underlying theoretical mechanisms would add depth and clarity to this area of study.
Selective Exposure and Cognitive Dissonance
While the studies reviewed earlier show evidence that selective exposure happens in the online news context, very few discuss why selective exposure occurs in any detail. The most commonly cited theoretical explanation for selective exposure is Festinger’s (1957) Theory of Cognitive Dissonance. According to Festinger, cognitive dissonance refers to a feeling of mental discomfort that arises when people are aware of inconsistencies between their attitudes and behaviors or between multiple attitudes that they hold. Moreover, Festinger shows that people are motivated to either reduce or avoid dissonance, which can happen by rationalizing a belief or behavior, diminishing its importance, or by selectively seeking information that confirms one’s belief or behavior. Indeed, the connection between cognitive dissonance and selective exposure has a long history in both the psychology and communication literatures (see Cotton, 1985; Donsbach, 2009; Harmon-Jones & Mills, 1999; Knobloch-Westerwick, 2014, for reviews).
There are several ways dissonance could arise from exposure to an attitude-challenging news source or information. Dissonance could arise from the behavior of supporting (e.g., giving viewership to) a source that advances an opposing ideology. Dissonance could also come from a lack of confidence in the accuracy of one’s beliefs after seeing attitude-challenging information, which is more likely to come from sources and/or information that uphold opposing ideologies. Selective exposure can reduce dissonance by helping people avoid (a) the behavior of supporting an attitude-challenging news source and (b) the psychologically uncomfortable situation of having their opinion called into question by disconfirming evidence or views. Moreover, selective exposure to attitude-consistent information can help reduce dissonance by reaffirming beliefs that people feel uncertain about (Cotton, 1985; Taber & Lodge, 2006).
Interestingly, most research on selective exposure using cognitive dissonance as its theoretical foundation assumes rather than tests the prediction that people feel dissonance when confronted with attitude-challenging news information (Cotton, 1985). In other words, past studies have not measured whether individuals actually feel dissonance when exposed to a counter-attitudinal message (e.g., Brannon et al., 2007; Garrett, 2009a, 2009b; Iyengar & Hahn, 2009; Jonas et al., 2001; Knobloch-Westerwick & Meng, 2009). Instead, these studies give people a choice among several news sources, and then assume anyone who did not select an attitude-challenging source did so because they knew they were likely to experience dissonance. Yet, without measuring cognitive dissonance, these studies cannot prove that people are motivated to seek out like-minded news information as a way of reducing or avoiding dissonance.
To address this problem, the present study examines differences in people’s experience of cognitive dissonance when exposed to attitude-challenging, attitude-consistent, and ideologically balanced information. Because cognitive dissonance theory predicts that selective exposure is a dissonance-avoidance strategy, it follows that people likely experience more cognitive dissonance when exposed to attitude-challenging information than when exposed to information that confirms their beliefs, and consequently, that people will select information sources that are attitude-consistent over those that challenge their attitudes. Therefore, the following hypotheses are proposed:
Credibility and Selective Exposure
While selective exposure research suggests that people have a preference for sources of information that share their political ideologies and confirm their perspectives (over attitudinally opposing sources), research on information credibility suggests that people may instead prefer balanced sources as more credible than unbalanced ones. This apparent contradiction in whether people favor attitude-consistent or politically balanced (i.e., “objective”) sources indicates a complicated relationship among credibility, source balance, and selective exposure. For example, it might indicate that although people find balanced sources more credible than unbalanced sources, they discount the importance of credibility when looking for political information, and thus select attitude-consistent sources. Alternatively, the contradiction might also be explained by research on biased information processing, as explained below.
Credibility, or the believability of a source or information, is typically defined as a function of a source’s expertise (e.g., credentials for discussing a topic, depth and breadth of knowledge) and trustworthiness (which is often indicated by the degree of “balance” of the information the source provides, for example, by covering multiple sides of an issue; Berlo, Lemert, & Mertz, 1969; Hovland, Janis, & Kelley, 1953). Credibility can be evaluated at the medium (e.g., Internet, television, newspaper), source (e.g., website, author), and message (e.g., news story) levels (Metzger, Flanagin, Eyal, Lemus, & McCann, 2003). In general research finds sources perceived as high in expertise are considered more credible than sources perceived as low in expertise, and balanced or objective sources (i.e., sources that do not support a particular ideology, as cued by name, affiliations, and the balance of political perspectives they present) are perceived as more credible than unbalanced or biased sources (i.e., sources that hold or advance a particular ideology; Allen, 1991; Pechmann, 1992; Zhao & Capella, 2008). Moreover, research finds that credibility judgments play a role during information search, with people being more likely to select a source that they think is credible than one they think is not credible (Kerstetter & Cho, 2004).
So, while credibility research indicates people would prefer balanced or objective news sources over attitude-consistent sources because they judge them as more credible than unbalanced sources, most studies of selective exposure within the context of partisan news have not examined this comparison because the typical research design used in that research focuses only on attitude-consistent versus attitude-challenging sources, without including balanced sources or information as a selection option (although there are exceptions, for example, Feldman, Stroud, Bimber, & Wojcieszak, 2013; Iyengar & Hahn, 2009). So, an interesting question is the extent to which people’s selection decisions will privilege balanced news sources over attitude-consistent news sources due to higher levels of perceived credibility. However, research on biased information processing suggests that the answer to this question may not be as simple as it might first appear.
Studies of bias perception in media coverage find that while people notice and devote greater attention to information that is antagonistic to their point of view, they perceive biased but attitude-congruent information as fair-minded (Gunther & Schmitt, 2004). Thus, while selective exposure to attitude-consistent sources could result from a drive to reduce and avoid dissonance alone, it could alternatively be due to people perceiving information from sources that agree with them is more impartial, and thus more credible—ironically in spite of the fact that those sources are indeed biased. Put another way, this suggests that people process biased information in biased ways: They notice bias when the source or message contradicts their attitudes and they use this as a strong negative credibility cue, but at the same time, they are rather blind to bias when the source or message is congruent with their attitudes and, in these cases, use attitude congruity as a positive credibility cue.
The idea that people may attribute higher levels of quality and fairness to biased but like-minded sources was first suggested in selective exposure research by Fischer, Jonas, Frey, and Schulz-Hardt (2005), and support for this as a possible explanation of selective exposure comes from Kahan and colleagues’ cultural cognition thesis (see Kahan, Braman, Cohen, Gastil, & Slovic, 2010; Kahan, Braman, Slovic, Gastil, & Cohen, 2009). The cultural cognition perspective combines elements of Wildavsky’s (1987) Cultural Theory of Preference Formation with research in psychology on cognitive heuristics. Wildavsky’s theory suggests that people filter information through their personal, cultural identities, and subsequently form opinions about that information. For example, when considering a proposed piece of legislation, people gauge the legislation’s ramifications against their own values, consider the opinions of others who have similar values, and evaluate the values of the legislation’s source. Kahan et al. (2010) argued that people do this because they tend to perceive like-minded sources as more honest, knowledgeable, and impartial than differently minded sources, independent of actual message quality. In other words, they use the source’s worldview as a shortcut to assess information credibility, making them more likely to value and choose like-minded sources. 2
In an application of the cultural cognition thesis to the study of information processing, Kahan et al. (2010) found that people rated attitude-consistent sources of information as more credible than counter-attitudinal sources. The cultural cognition account is also consistent with Meyer, Marchionni, and Miller (2010) and with Oyedeji (2010), who found that credibility and attitude-consistency with a news source were positively linked, and with the results of Fischer et al.’s work (Fischer et al., 2005; Fischer et al., 2008) on selective exposure arguing that news consumers may use source balance (i.e., whether they feel a source is not politically biased) as a mental shortcut when assessing credibility.
At the same time, research on the hostile media phenomenon suggests that individuals may attribute antagonistic biases even to unbiased news sources, especially when they are highly involved in the issue being covered (Christen, Kannaovakun, & Gunther, 2002; Giner-Sorolla & Chaiken, 1994). For example, in a study on perceptions of bias in media coverage of the 1982 Beirut Massacre, both pro-Israeli and pro-Arab viewers perceived antagonistic biases from the same balanced news coverage (Vallone et al., 1985). This demonstrates that “objective” or balanced sources are not always perceived by news consumers as unbiased.
Scholars have only recently begun examining the role of credibility perceptions in partisan selective exposure. Stroud and Lee (2013) found that credibility perceptions for cable news channels mediated the relationship between political party identification and exposure to partisan news. Johnson and Kaye (2013) found a similar relationship among credibility, political predispositions, and exposure to news blogs: Blog users were likely to expose to biased sources that shared their political point of view.
In sum, the foregoing research indicates that individuals may find biased, yet attitude-congruent news sources more credible than balanced or opinion-challenging sources because they perceive attitude-congruent sources as more impartial, and thus they may be likely to seek out and rely on like-minded news sources. As such, credibility offers a new theoretical explanation for selective exposure that has only recently begun to be explored. To examine the role of credibility perceptions in selective exposure, the following hypotheses are advanced:
As discussed previously, information consumers evaluate credibility at both the source and message levels: They may use the political bias of a news source (e.g., a website such as FoxNews.com) as well as cues inherent in the information provided on a news source (e.g., aspects of a particular news story such as whether it presents politically balanced information; see Metzger et al., 2003). However, it is unclear how source bias and story balance may interact to impact credibility perceptions. Therefore, the following research question is proposed:
Method
Participants
An online experiment was used to examine the effects of perceptions of online news credibility and cognitive dissonance on selective exposure. Data for this study were collected by the professional survey research firm Knowledge Networks (now GfK), which maintains a probability-based panel of participants that matches the U.S. population. Panel recruitment is done via address-based sampling. Potential panel members are sent an invitation to join the panel in the mail, and are allowed to sign up by calling, emailing, or returning an application via mail. Panel members are expected to complete one 15- to 20-minute study per week. They are entered in prize drawings as an incentive to complete studies.
A total of 2,146 adults living in the United States participated in this study: 48.6% (1,043) participants were male and 51.4% (1,103) were female; 35.9% (771) identified as Democrat, 27.8% (596) identified as Republican, 24.4% (524) identified as Independent, 5.5% (117) identified as Tea Party members, 0.7% (16) identified as Green Party members, and 4.6% (99) identified as “other”; participants ranged in age from 18 to 94 years (M = 49.25 years, SD = 16.03); 65.4% (1,403) participants were White, 10.1% (217) were Black, 10.4% (224) were Hispanic, 11.6% (249) identified as “Other, non-Hispanic,” and 2.5% (53) identified as mixed-race.
The experiment employed a 3 (source balance: conservative, liberal, or unbiased news source) × 3 (story balance: conservative, liberal, or balanced news stories) design. Conservative stories were created by providing quotes from Republican politicians (e.g., Mike Huckabee), emphasizing conservative stances on issues (i.e., pro-life, anti-gay marriage) and criticizing liberal viewpoints. Likewise, liberal stories were created by using quotes from Democrat politicians (e.g., Nancy Pelosi), emphasizing liberal stances on issues and criticizing conservative viewpoints. Balanced stories provided quotes about liberal and conservative viewpoints without taking a side. The news sources were all screenshots of websites (see Figure 1 for an example), and included FoxNews.com (conservatively slanted traditional source), NPR.com (liberally slanted traditional source), CNN.com (balanced traditional source), LiberalOasis.com (liberal blog), RedState.com (conservative blog), and TheModerateVoice.com (balanced blog). The news sources were categorized as liberal, conservative, or balanced based on study participants’ own perceptions, as detailed in “Measures” section.
Example stimulus.
The news sources were manipulated to include stories on two issues: the legalization of gay marriage in Washington, D.C., and budget cuts to Planned Parenthood programs because of the organization’s support of abortion. These stories were manipulated as mentioned above to be either balanced (present opposing views on the issue) or unbalanced (emphasize either the liberal or conservative perspective on the issue).
Each participant in the experiment saw only one news source and read only one story from that source, selected randomly from among the experimental stimuli. After viewing the source and story (i.e., the news website), participants answered questions pertaining to the degree of cognitive dissonance they felt while reading the story, their evaluations of the credibility of the news sources, and other variables, including selective exposure, as described next.
Measures
The dependent and independent variables in this study, as well as several control and demographic variables were measured via several items. The independent variables were pilot tested to ensure that people would correctly perceive differences based on the source type, source balance, and story balance manipulations. Results of these pretests are described below where relevant. In addition, manipulation checks were employed within the main study.
Source balance
Source balance was conceptualized in terms of a news source’s political conservatism, liberalism, or objectivity. Unbalanced sources are those that are known to espouse a certain political perspective, whereas objectivity was operationalized in terms of sources that are recognized to have no clear affiliation with a conservative or liberal political ideology. The study included both traditional news outlets and blogs, although differences between them were not examined in this study. Sources in the conservative conditions were Fox News and Red State (a leading conservative news blog). Sources in the liberal conditions were NPR and Liberal Oasis (a popular liberal news blog). Sources in the balanced conditions were CNN and The Moderate Voice (self-described as one of the most respected centrist news blogs). All sources were pilot tested with a separate sample to assess whether people associated the correct political bias with each source, or lack of bias for the balanced sources. Results of the pilot test were significant and consistent with expectations.
In addition to the pilot study, participants’ perception of source balance was also measured in the main experiment directly on a 5-point Likert-type scale (1 = conservative, 5 = liberal). As expected, liberal sources (M = 3.76, SD = 1.31) were rated significantly higher on the scale than balanced sources (M = 3.24, SD = 1.24), which were themselves rated significantly higher on the scale than the conservative sources (M = 2.28, SD = 1.16). 1 In this way, participants’ perception of source balance was used as a manipulation check to ensure that participants understood correctly the political bias of the source they saw. Anyone who failed the manipulation check was excluded from further analysis. 3
Story balance
As with source balance, story balance was both manipulated and measured in this study. The news stories were written as either unbalanced or balanced by either (a) emphasizing liberal views on the issue and presenting those views as superior to conservative views, (b) emphasizing conservative views on the issue and presenting those views as superior to liberal views, or (c) presenting both liberal and conservative views on the issue equally and as equally valid. Stories in categories (a) and (b) were coded as unbalanced, whereas stories in category (c) were coded as balanced.
Participants’ perceptions of story balance were measured on a Likert-type scale (1 = conservative, 7 = liberal). Analyses showed that liberal stories (M = 5.38, SD = 1.36) were rated significantly higher on the scale than either balanced stories (M = 3.95, SD = 1.47) or conservative stories (M = 2.42, SD = 1.68). This variable was used as a check on the story balance manipulation in order to ensure that participants understood correctly the bias of the story they read. Anyone who failed the manipulation check was excluded from further analysis.
Political ideology
Political ideology was measured by asking participants which political party they most identify with (Democratic, Republican, Green, Tea, Independent, or other). Those who identified with the Democratic or Green party were coded as liberal, participants who identified with the Republican or Tea party were coded as conservative, and those who identified as Independent or “other” were coded into their own groups.
Attitude-consistency of source
To operationalize attitude-consistency between source and participant (i.e., whether a participant saw an attitude-challenging, attitude-consistent, or balanced source), a variable was created based on each participant’s political party identification and the source balance manipulation. People who identified as members of the Democratic or Green Party who saw a liberal source were categorized as having seen an “attitude-consistent” source, as were people who identified as members of the Republican or Tea Party who saw one of the conservative sources. People who identified with a liberal political party and who saw a conservative source were coded as having seen an “attitude-challenging” source, as were people who identified with a conservative party and who saw a liberal source. Liberal and conservative participants who saw a balanced source were coded into a third category, as having seen a source that was neither attitude-consistent nor attitude-challenging.
Independents could not be reliably coded as having seen an attitude-consistent or attitude-challenging source, so two separate categories were created such that Independents were coded as having seen either a balanced news source or an unbalanced news source.
Attitude-consistency of story
To determine whether a participant saw an attitude-challenging, attitude-consistent, or balanced story, a variable was created based on each participant’s score on an issue-support scale. Participants were asked for their opinion regarding gay marriage (“I believe that gay men and women should be allowed to marry”) and Planned Parenthood (“I believe that Planned Parenthood should continue to receive federal funding”) on a 5-point Likert-type scale (1 = strongly disagree, 5 = strongly agree). Participants were coded as seeing an attitude-consistent news story if they (a) indicated a 1 or 2 on the issue-support scale and saw a conservative story or (b) indicated a 4 or 5 on the issue-support scale and saw a liberal story. Participants were coded as seeing an attitude-challenging story if they (a) indicated a 1 or 2 on the issue-support scale and saw a liberal story or (b) indicated a 4 or 5 on the issue-support scale and saw a conservative story. All participants who saw a balanced story were coded as having seen a story that was neither attitude-consistent nor attitude-challenging. As before, Independents could not be coded reliably as having seen an attitude-consistent or attitude-challenging story, so two separate categories were created: Independents were coded as having seen either a balanced story or an unbalanced story.
Credibility perceptions
Story and source credibility were measured using items developed by Flanagin and Metzger (2000). Story credibility was defined in terms of the completeness (“How complete is the information presented in this news story?”), accuracy (“How accurate do you find the information to be?”), bias (“How unbiased do you find the information to be?”), trustworthiness (“How trustworthy do you find the information to be?”), and credibility (“How credible do you find the information to be?”) of the information presented in the news story. Each of these dimensions was measured on a 5-point Likert-type scale. The scale was constructed by averaging across scores on the five dimensions. Cronbach’s alpha reliability coefficient for the scale was .87.
Source credibility was conceptualized as participants’ perceptions of a news source’s bias, professionalism, and trustworthiness, and was assessed using four items, each on a 5-point Likert-type scale. Items included “How biased do you find this website to be?” (reverse coded) “How much do you trust this website?” “How professional do you find this website to be?” and “How credible do you feel this website is?” The scale was constructed by averaging across scores on the four items. Cronbach’s alpha for the scale was .82.
Cognitive dissonance
As mentioned earlier, past research has mostly assumed rather than actually measured the experience of cognitive dissonance as a result of exposure to attitude-inconsistent information (Cotton, 1985; Devine, Tauer, Barron, Elliot, & Vance, 1999). Indeed, besides measuring general arousal levels after exposure to information, the only existing measure of cognitive dissonance is the “dissonance thermometer” developed by Elliot and Devine (1994). Although useful, the dissonance thermometer is an indirect measure in that it gauges self-reported affect that is theoretically associated with cognitive dissonance (discomfort). But cognitive dissonance has both affective and cognitive dimensions (Martinie, Milland, & Olive, 2013). Moreover, the dissonance thermometer has only been applied in the situation of dissonance aroused within the counter-attitudinal advocacy paradigm (i.e., writing a counter-attitudinal essay), rather than by exposure to news information, which is a quite different psychological context.
Because there are no existing measures of cognitive dissonance stemming from exposure to attitude-challenging news sources/information and that tap into the cognitive and affective dimensions of cognitive dissonance, a new scale was developed to assess participants’ experience of dissonance while reading the news site and story. Consistent with past self-report measures of the dissonance people experience when forced to advocate for a position they disagree with (Elliot & Devine, 1994), dissonance was conceptualized as mental discomfort arising from inconsistency between two beliefs (i.e., in our case a participant’s belief on an issue and the belief expressed in a news story about that issue) or between a belief and an action (i.e., identification with a political party and support for a news source that holds an opposing political ideology). Dissonance was measured by averaging responses across nine 5-point Likert-type scale items that were designed to tap into the cognitive or emotional dimensions of cognitive dissonance (Martinie et al., 2013). Items included, “I regret reading this news story,” “This news source makes me uncomfortable,” “I disliked reading this story because it challenged my beliefs,” “I agreed with the stance taken in the article” (reverse coded), “I felt uncomfortable while reading this news story,” “This story made me question my own beliefs about the issue [gay marriage or abortion],” I enjoyed reading this news story” (reverse coded), “I would feel uncomfortable supporting this news source,” and “I like this news source” (reverse coded). Cronbach’s alpha for the scale was .82.
Selective exposure
Selective exposure was conceptualized as the likelihood that participants would use a news source again in the future. It was measured by asking participants, “How likely would you be to select this website rather than another source for news information in the future?” (5-point Likert-type scale, 1 = not at all likely, 5 = very likely). Although this is not a direct behavioral measure, meta-analytic research suggests that behavioral intent is a reliable predictor of actual behavior (Sheeran, 2002). While a direct behavioral measure of news source selection would have been preferred, the present selective exposure measure was necessary because most participants would be likely to select attitude-consistent sources if given a choice (Knobloch-Westerwick & Meng, 2009), and this would have precluded the ability to compare the amount of dissonance invoked by attitude-consistent versus attitude-challenging sources, which is key to this study. Similar “forced exposure” designs have a long history in selective exposure research (see Fischer & Greitemeyer, 2010) and have been employed in recent studies as well (e.g., Arceneaux, Johnson, & Murphy, 2012; Coe et al., 2008; Feldman et al., 2013). Also, measuring selective exposure via self-reported likelihood of news consumption (rather than via behavioral observation) is common in the literature (e.g., Garrett, 2009b; Johnson et al., 2009; Kobayashi & Ikeda, 2009; Melican & Dixon, 2008; Stroud, 2008).
Source familiarity
Research indicates that familiarity with a news source impacts people’s perceptions of that source in biased ways. For example, frequent users of particular news sources, including mainstream news and blogs, rate those sources as more credible compared with news sources that individuals use less frequently (Johnson & Kaye, 2009). Thus, the present study controls for source familiarity in all relevant analyses. Participants were asked how familiar they were with the news website they viewed in the study, measured on a 5-point Likert-type scale (1 = not at all familiar, 5 = very familiar).
Results
Hypothesis 1 was tested via an ANCOVA with source attitude-consistency (i.e., attitude-consistent, attitude-challenging, and balanced sources) as the independent variable and cognitive dissonance as a dependent variable, controlling for source familiarity. Hypothesis 1 predicted that participants would report higher levels of cognitive dissonance when exposed to attitude-challenging news sources than when exposed to attitude-consistent or balanced news sources. As mentioned earlier, participants were excluded from analysis if they failed the source balance manipulation check. The analysis revealed a significant main effect for source attitude-consistency on cognitive dissonance, F(2, 689) = 50.51, p < .001, ηp2= .13. Post hoc tests showed that viewing attitude-challenging news sources (M = 2.92, SE = 0.04) caused significantly greater cognitive dissonance than did attitude-consistent (M = 2.36, SE = 0.04) or balanced (M = 2.58, SE = 0.04) news sources (see Table 1). In addition, balanced news sources caused significantly more cognitive dissonance than attitude-consistent sources. Hypothesis 1 was supported.
Mean Differences on Dependent Variables Among Partisans Viewing Attitude-Challenging, Attitude-Consistent, and Balanced News Sources.
Note. Row means with common subscripts do not differ significantly from one another.
p < .001.
Hypotheses 2 and 5 were tested using ANCOVA with source attitude-consistency (i.e., attitude-consistent, attitude-challenging, and balanced sources) as the independent variable and selective exposure as the dependent variable. Again, participants were excluded from analysis if they failed the source balance manipulation check. The hypotheses predicted that participants would be more likely to select attitude-consistent news sources than either attitude-challenging news sources (Hypothesis 2) or balanced news sources (Hypothesis 5). An ANCOVA controlling for news source familiarity revealed a significant main effect for source attitude-consistency on selective exposure, F(3, 910) = 39.43, p < .001, ηp2 = .12. Post hoc follow-up tests showed that liberals and conservatives who saw an attitude-consistent source (M = 2.52, SE = 0.07) were significantly more likely to say they would return to the news source in the future compared with liberals and conservatives who saw an attitude-challenging source (M = 1.70, SE = 0.06). However, there was no significant difference between liberals and conservatives who saw an attitude-consistent source and those who saw a balanced source (M = 2.34, SE = 0.07) in terms of selective exposure (see Table 1). Thus, the data support Hypothesis 2 but do not support Hypothesis 5.
Although not directly addressed in Hypothesis 2, the results for political Independents offer further support for this hypothesis if we consider that a balanced source is “attitudinally consistent” for Independents, whose stance is neither liberal nor conservative, and indeed may include views from both ends of the political spectrum. The data show that independents who saw a balanced source (M = 2.38, SE = 0.10) were more likely to report they would return to the news source than independents who saw an unbalanced source (M = 1.77, SE = 0.07; Table 2). Comparing this with the data reported in the previous paragraph (and in Table 1), Independents who saw a balanced source were also more likely to say they would return to it than liberals and conservatives who saw an attitude-challenging source, and liberals and conservatives who saw an attitude-consistent source were more likely to return to it than independents who saw an unbalanced source. Finally, liberals and conservatives who saw a balanced source said they were more likely to return to it than did liberals and conservatives who saw an attitude-challenging source and independents who saw an unbalanced source. These results are consistent with Hypothesis 2 in that they show a preference for attitudinally consistent sources of news.
Follow-Up Analyses With Nonpartisans and Issue-Neutrals, Mean Differences on Selective Exposure and Credibility.
Note. All row means are significantly different at p < .001.
Hypothesis 3 predicted that participants would rate attitude-consistent sources higher in credibility than attitude-challenging or balanced sources. Participants were again excluded from analysis if they failed the source balance manipulation check. An ANCOVA was performed to test Hypothesis 3, with source attitude-consistency as the independent variable and source credibility as the dependent variable, controlling for source familiarity. Analyses revealed a significant main effect for source attitude-consistency on source credibility, F(3, 939) = 68.76, p < .001, ηp2 = .18. Pairwise comparisons showed that liberals and conservatives who saw an attitude-consistent source (M = 3.05, SE = 0.05) perceived it as significantly more credible than those who saw attitude-challenging source (M = 2.31, SE = 0.05). In addition, liberals and conservatives who saw a balanced source (M = 3.14, SE = 0.06) judged it as significantly more credible than an attitude-challenging source, but not more credible that an attitude-consistent source (see Table 1). Also, as shown in Table 2, independents who saw a balanced source (M = 2.98, SE = 0.08) judged it as significantly more credible than independents who saw an unbalanced source (M = 2.38, SE = 0.06) as well as liberals and conservatives who saw an attitude-challenging source. Thus, Hypothesis 3 was partially supported: Individuals do judge attitude-consistent sources as significantly more credible than attitude-challenging sources, but there was not a significant difference between the credibility of attitude-consistent sources and balanced sources.
Hypothesis 4 predicted that participants would rate attitude-consistent news stories as more credible than attitude-challenging or balanced news stories. An ANCOVA was performed with story attitude-consistency as the independent variable and story credibility as the dependent variable, controlling for source familiarity, and excluding any participant if they failed the story balance manipulation check. Analyses showed a significant main effect for story attitude-consistency on story credibility, F(4, 242) = 11.49, p < .001, ηp2 = .16. As shown in Table 1, people who saw an attitude-consistent story (M = 3.00, SE = 0.10) perceived it as significantly more credible than people who saw an attitude-challenging story (M = 2.32, SE = 0.08) but no different from balanced stories (M = 3.19, SE = 0.25). In addition, people with a liberal or conservative issue stance rated balanced stories (M = 3.06, SE = 0.10) as significantly more credible than attitude-challenging stories (M = 2.32, SE = 0.08). Moreover, Table 2 shows people who were neutral on the issue presented in the story judged balanced stories (M = 3.32, SE = 0.40) as significantly more credible than unbalanced stories (M = 2.48, SE = 0.19). Hypothesis 4 was thus partially supported: Attitude-consistent stories were judged as more credible than attitude-challenging stories, but they were not judged as significantly more credible than balanced stories.
Research Question 1 asked about the interaction between the attitude-consistency of the source and the attitude-consistency of the story on participants’ judgment of the source’s credibility. To answer this question, which gets to the issue of whether people rate attitude-consistent sources that also display attitude-consistent information at the story level as more credible than other combinations of attitudinally consistent, inconsistent, or balanced sources and stories, participants were first coded into seven categories: (a) Liberals or conservatives who saw an attitude-consistent story from an attitude-consistent source were coded as seeing “very consistent” information. (b) Liberals or conservatives who saw an attitude-consistent story from a balanced source or a balanced story from an attitude-consistent source were coded as seeing “somewhat consistent” information. Independents who saw an unbalanced story from a balanced source or a balanced story from an unbalanced source were also coded as seeing “somewhat consistent” information. (c) Liberals and conservatives who saw a balanced story from a balanced source were coded as seeing “very balanced” information. (d) Liberals and conservatives who saw an attitude-challenging story from a balanced source or a balanced story from an attitude-challenging source were coded as seeing “somewhat challenging” information. (e) Liberals and conservatives who saw an attitude-challenging story from an attitude-challenging source were coded as seeing “very challenging” information. (f) Independents who saw an unbalanced story from an unbalanced source were coded as seeing “very unbalanced” information. (g) Finally, independents who saw a balanced story from a balanced source were coded as seeing “very balanced” information, but were analyzed separately from liberals and conservatives who saw “very balanced” information, because it could not be determined whether they fit better in the “very consistent” or “very balanced” category.
To answer Research Question 1, an ANCOVA was run with the seven-category source-by-story-attitude-consistency variable as the independent variable, source credibility as the dependent variable, and source familiarity as a control. This approach was taken because a conventional mathematical interaction term could not correctly classify the data into the relevant categories because of the inclusion of the political Independents. This approach offers an alternative way to determine whether attitude-consistency at both the source and story levels simultaneously affect credibility perceptions, as posed by Research Question 1. As before, participants were excluded from analysis if they failed either the source balance or story balance manipulation check.
A significant effect was found for source-and-story attitude-consistency on source credibility, F(6, 547) = 26.31, p < .001, ηp2 = .22. As shown in Table 3, the most credible source-and-story combinations were found for liberals and conservatives seeing balanced stories from balanced sources, followed by individuals who saw somewhat attitude-consistent information (i.e., attitude-consistent source with balanced story, or vice versa), independents who saw “very balanced” information, liberals and conservatives who saw “very consistent” information, individuals who saw “somewhat challenging” information (i.e., a balanced story from attitude-challenging source, or vice versa), independents who saw “very unbalanced” information, and liberals and conservatives who saw “very challenging” information. Among these, “very challenging” and “very unbalanced” information were perceived as significantly less credible than all other types of information. Thus, the combination of viewing a source with a story that challenges attitudes (by opposing partisans’ attitudes or by presenting biased information to political centrists) resulted in the lowest credibility scores.
Effect of Interaction Between Attitude-Consistency of Source and Story on Credibility Perceptions for Partisans and Nonpartisans, Means Shown in Descending Order.
In summary, looking across all the analyses, it was found that news consumers experience greater cognitive dissonance when viewing news from attitude-challenging sources, rate attitude-consistent and balanced sources and stories as more credible than attitude-challenging sources and stories, and report greater likelihood of selecting attitude-consistent sources over attitude-challenging sources, but not balanced sources. In addition, the degree of attitude-consistency across both source and story appear to work simultaneously to affect news credibility judgments. The theoretical implications of these results are discussed next.
Discussion
This study advances research on cognitive dissonance theory, selective exposure, and credibility in several ways. First, it challenges the viability of cognitive dissonance as an explanatory mechanism for selective exposure. Second, it tests whether credibility judgments are an (perhaps superior) explanation of selective exposure. Third, it suggests that news consumers—including partisans—are likely to use unbiased sources, which tempers fears about the negative effects of selective exposure on democracy. Finally, it sheds new light on how source and story bias may interact to impact credibility perceptions. Also, the sample for this study was selected at random and was more demographically and cognitively diverse than the typical laboratory experiment, which may enhance the generalizability of the results reported here.
This study represents the first attempt to actually measure felt cognitive dissonance in a mass media selective exposure study. The results of this study offer some support for the cognitive dissonance explanation of selective exposure, and yet conflict with it in other ways. Consistent with the predictions of cognitive dissonance theory, news consumers experience more cognitive dissonance when exposed to attitude-challenging news sources than attitude-consistent sources. Also as predicted, participants reported being more likely to return to attitude-consistent sources than to attitude-challenging sources. However, although individuals reported experiencing more cognitive dissonance when exposed to unbiased news sources than to attitude-consistent news sources, they are equally likely to select these two source types. If dissonance avoidance is the true motivator for source selection, participants should have reported being more likely to return to attitude-consistent sources than to unbiased sources. 4
Further evidence for problems with the dissonance explanation for selective exposure is that across all conditions in the study, cognitive dissonance was quite low, typically reported on the lower side of the middle of a 5-point scale (i.e., most participants fell between disagree and neither agree nor disagree when asked if they felt dissonance while reading the story) for all sources and stories, and yet we detected statistically significant differences in selective exposure across conditions. This suggests that cognitive dissonance is not a very powerful mechanism driving selective exposure, at least in the context of selective exposure to partisan news.
Selective exposure patterns fell more in line with credibility perceptions than with reports of cognitive dissonance. While participants judged attitude-challenging sources as less credible than attitude-consistent sources, they judged attitude-consistent and unbiased sources as equally credible. Likewise, participants judged attitude-challenging news stories low in credibility, but perceived attitude-consistent and balanced news stories as equally credible. The same pattern appeared in self-reported selective exposure behavior: People in this study were unlikely to rely on attitude-challenging sources, but were equally likely to rely on attitude-consistent and balanced sources. This provides evidence supporting the notion that people process attitude-confirming, yet biased information as credible, and that these credibility judgments predict selective exposure behavior more consistently than does cognitive dissonance.
These findings have other important implications for selective exposure research. Similar to other recent studies on selective exposure to partisan news, this study found that partisans are not likely to rely on sources that counter their worldviews, preferring instead sources that match their own political perspectives (e.g., Stroud, 2008, 2011). However, the results also show that even partisans appear to have a healthy respect for neutral sources as well, which may temper fears about the implications of selective exposure. If most people are willing to rely on unbiased sources in addition to attitude-consistent sources, as the current study suggests, this may minimize such effects of selective exposure discussed in the literature as attitude polarization and out-group hostility, a less-than-fully informed citizenry, as well as political apathy, enmity, and deadlock (Bennett & Iyengar, 2008; Christen & Gunther, 2003; Dilliplane, 2011; Levendusky, 2013; Stroud, 2010).
The present study also reveals a possible interaction between the balance of the source and the balance of the story itself on news consumers’ credibility judgments, confirming past research that suggests credibility is simultaneously evaluated at both source and message levels (Metzger et al., 2003). Overall, balanced sources that provide balanced information were uniformly judged as the most credible among all study participants. In a way, these findings appear to conflict with past research on the hostile media phenomenon that finds balanced (unbiased) stories may be perceived by partisans as biased (Vallone et al., 1985). In the current study, balanced sources and stories did not take a credibility “hit” by virtue of partisan participants feeling they were biased. Our results may differ from hostile media effect studies due to differences in the importance of the issues to participants (who were highly involved with the issue in the Vallone et al. study, whereas our participants may have been less involved with the issues presented in this study) or the variables measured (this study examined credibility, whereas hostile media studies examine bias exclusively).
At the same time, this study’s findings conflict with some past credibility research that suggests information consumers reject unbalanced information as not credible regardless of its attitude-consistency (see Allen, 1991). The findings of the current study, in conjunction with recent research on selective exposure and newer theories about biased information processing, indicate that this is not the case. People appear to judge attitude-consistent sources and balanced sources as comparably, if not equally, credible, and both credibility and selective exposure researchers should account for this in future research.
Limitations and Suggestions for Future Research
While this study makes some important contributions to the literatures on selective exposure and credibility, like all research it has some limitations. One criticism comes from the measure of cognitive dissonance. Festinger (1964) stated that cognitive dissonance occurs only after people have committed to an attitude or behavior and then put themselves in a situation where the validity of their position or behavior is challenged. There is an element of choice involved in cognitive dissonance theory that was not present in the present study. For example, if participants had chosen to look at an attitude-challenging instead of an attitude-consistent source (rather than being shown a source) they might have experienced more dissonance.
The operationalization of selective exposure could be similarly criticized, as participants were presented with news stories rather than having an opportunity to select the stories themselves. It is also possible that because selective exposure was measured as a self-reported future action, participants reported the socially desirable response (i.e., being likely to return to a balance source in the future), and this accounts for the equality in selective exposure between balanced and attitude-consistent sources. As explained in the “Method” section, both forced exposure and self-reports of selective exposure are not uncommon in the literature, but future research designs that include the ability for respondents to choose which sources and stories they pay attention to and then measure cognitive dissonance are necessary to establish further evidence of cognitive dissonance as an explanation for selective exposure in news. That said, a significant contribution of this study is that it is the first to provide a means of measuring cognitive dissonance in news-seeking situations.
In light of recent evidence that news consumers have a tendency to seek out information from like-minded sources, longitudinal research needs to be done to examine the effect of selective exposure on attitude-reinforcement and political polarization over time. To date, very few studies have examined the long-term effects of selective exposure on political polarization but Stroud (2010) found that liberals and conservatives who rely heavily on attitude-consistent sources become more entrenched in their views and more hostile toward opposing ideologies over time. This is an important finding, and scholars should try to replicate it.
Longitudinal data may also help to understand better the relationship between credibility and selective exposure. We have argued here that perceptions of the credibility of a news source increases selective exposure to that source, but it could be that selective exposure to a source leads to increased perceptions of its credibility. The cross-sectional nature of our data limits our ability to examine this issue. More research also needs to be done to explore the connection between credibility evaluations at the source and story levels and selective exposure in light of the unexpected finding that people find balanced stories from balanced sources as the most credible type of information.
It is also unclear whether source or message credibility plays a larger role in selective exposure. The answer to this question most likely depends on search motivation. Lay epistemics theory (Kruglanski, 1989) and the theory of motivated reasoning (Kunda, 1990) suggest that the rigor of people’s information searches is guided by whether they desire accurate (objective) or directional (biased) information. People who desire accurate information will perform a more rigorous search for information and will select more diverse information than people who want directional information (i.e., information that confirms a preferred position). While some research has examined third variables that impact selective exposure behavior, such as the amount of time people have to search for information (Fischer et al., 2005; Fischer et al., 2008; Kim, 2007), more research is needed to understand the search circumstances under which selective exposure is most likely to occur and how the search process impacts people’s judgments of the credibility of balanced versus attitude-consistent information.
Finally, this study indicates that cognitive dissonance and credibility both play a role in selective exposure, but it does not examine the relationship between cognitive dissonance and credibility directly. Many interesting questions await further research, including, for example, when does credibility judgment affect cognitive dissonance, and vice versa? Will information that causes cognitive dissonance (because it challenges a preferred position or comes from an attitude-challenging source) be judged as low in credibility as a dissonance-reduction strategy? Do people experience less cognitive dissonance when attitude-challenging information comes from a credible source? And, will attitude-consistent information from a source that is low in credibility cause cognitive dissonance?
Conclusion
Although this study provides some evidence of selective exposure behavior in news consumption, it suggests several reasons for optimism as well. People do seem to prefer attitude-consistent sources over attitude-challenging ones, but they are also likely to use balanced sources. While this does not completely put to rest fears that the United States is becoming more polarized, it does suggest that these fears may be overstated. The mere presence of attitude-consistent sources does not necessarily preclude people from seeking out attitude-challenging information, and it certainly does not seem to stop people from using the many balanced news sources available in the contemporary media environment.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was generously supported by the John D. and Catherine T. MacArthur Foundation.