Summarize the following article.

https://doi

.

org/10.1177/00936502177195

96

Communication Research
1 –21

© The Author(s) 20

17

Reprints and permissions:

sagepub.com/journalsPermissions.nav
DOI: 10.1177/0093650217719596

journals.sagepub.com/home/crx

Article

Confirmation Bias, Ingroup
Bias, and Negativity Bias in
Selective Exposure to
Political Information

Silvia Knobloch-Westerwick1, Cornelia Mothes1,
and Nick Polavin1

Abstract
Selective reading of political online information was examined based on cognitive
dissonance, social identity, and news values frameworks. Online reports were
displayed to 156 Americans while selective exposure was tracked. The news articles
that participants chose from were either conservative or liberal and also either
positive or negative regarding American political policies. In addition, information
processing styles (cognitive reflection and need-for-cognition) were measured. Results
revealed confirmation and negativity biases, per cognitive dissonance and news values,
but did not corroborate the hypothesis derived from social identity theory. Greater
cognitive reflection, greater need-for-cognition, and worse affective state fostered the
confirmation bias; stronger social comparison tendency reduced the negativity bias.

Keywords
selective exposure, confirmation bias, social identity, news values, political information

Audience members predominantly select what media messages they attend to and thus
shape their individual information environments. Biases in political media coverage
have long raised concerns among scholars (e.g., D’Alessio & Allen, 2000), but media
users play a pivotal role in creating their own potentially biased information environ-
ment by selectively attending political messages. Specifically in the Internet age,
where information selection has become particularly convenient due to online access,

1The Ohio State University, Columbus, OH, USA

Corresponding Author:
Silvia Knobloch-Westerwick, The Ohio State University, 154 N Oval Mall, Columbus, OH 43210, USA.
Email: knobloch-westerwick.1@osu.edu

719596CRXXXX10.1177/0093650217719596Communication ResearchKnobloch-Westerwick et al.
research-article2017

https://us.sagepub.com/en-us/journals-permissions

https://journals.sagepub.com/home/crx

mailto:knobloch-westerwick.1@osu.edu

http://crossmark.crossref.org/dialog/?doi=10.1177%2F0093650217719596&domain=pdf&date_stamp=2017-07-18

2 Communication Research 00(0)

many scholars are concerned with factors that may result in biased message consump-
tion (e.g., Bennett & Iyengar, 2008). Festinger’s (1957) theory of cognitive dissonance
guided numerous studies that investigated whether media users prefer attitude-consis-
tent content over attitude-discrepant messages (Donsbach, 2009). Another theory by
Festinger (1954), social comparison theory, inspired social identity theory (Tajfel &
Turner, 1979), which has guided some recent research into news selection and pro-
cessing (e.g., Iyengar, Sood, & Lelkes, 2012; Stroud, Muddiman, & Lee, 2014). Per
social identity theory, media consumers should prefer messages that present their
ingroup in a favorable light, even though classic news factors research suggests that
news consumers are generally more attracted to negative news.

In the following, these three theoretical perspectives will be reviewed to derive
hypotheses. What information recipients choose to attend to is of central interest to
this investigation. The term selective exposure is used here simply to denote that indi-
viduals choose from messages and exhibit preferences in those selections, which may
take the form of a confirmation bias but can also demonstrate other leanings. An
empirical study will examine the confirmation bias, the ingroup bias, and the negativ-
ity bias simultaneously in a selective exposure study in which American participants
browsed online messages that aligned either with liberal or conservative stances and
that either praised or critiqued the United States in comparison with another country.

Confirmation Bias: A Cognitive Dissonance Perspective

The postulation that individuals prefer messages that align preexisting attitudes over
those messages that challenge them (Festinger, 1957) has not found consistent support
throughout decades of research (Donsbach, 2009). Specific meta-analytic examination
with focus on postdecision selective-exposure in controlled settings did indeed sup-
port Festinger’s (1957) propositions (D’Alessio & Allen, 2002). A recent surge of
empirical investigations in particular for political news and information, with an
emphasis on online contexts, has yielded more and more consensus that information
users favor attitude-consistent messages (e.g., review by Knobloch-Westerwick,
2015). While debate exists on whether the avoidance of attitude-discrepant messages
is as strong as the preference for attitude-consistent content (Garrett, 2009) and how
the confirmation bias may foster political polarization (e.g., Iyengar et al., 2012;
Stroud, 2010), the current view is that information users generally approach political
messages with a confirmation bias. Hence, the present work also expects to find evi-
dence of this bias.

Hypothesis 1 (H1): Selective exposure is greater for political messages that align
with preexisting political attitudes than selective exposure to messages that chal-
lenge them (confirmation bias).

Given the concern that citizens may restrict themselves largely to attitude-consis-
tent information with negative consequences for political discourse and polarization
(Bennett & Iyengar, 2008; Knobloch-Westerwick, 2012), stable and situational factors
that might moderate the extent of this confirmation bias need to be examined. Stable

Knobloch-Westerwick et al. 3

interindividual differences have occasionally been examined (dogmatism, authoritari-
anism, chronic anxiety, and repression/sensitization; see reviews by Hart et al., 2009;
Smith, Fabrigar, & Norris, 2008). Recent work suggests that information processing
styles and the way in which attitude-consistent versus attitude-discrepant exposure is
operationalized may affect the confirmation bias and its consequences for attitudes:
Specifically, Taber and Lodge (2006) proposed that barring attitude-discrepant influ-
ences takes cognitive efforts. Simply put, selecting information that converges with
preexisting views involves more cognitive engagement and preliminary processing of
cues about the content of available messages than relatively mindless message choices.
Thus, with the second hypothesis, the present work will also examine whether these
stable interindividual differences in information processing styles influence the con-
firmation bias—also in light of the fact that the ample related research has only rarely
considered such traits (Hart et al., 2009; Smith et al., 2008). Research by Kahan (2013)
suggests that the tendency to override first intuitive conclusions in favor of more
thoughtful processing as measured by Frederick’s (2005) cognitive reflection (CR)
test favors even greater motivated processing along the lines of a confirmation bias, in
line with Taber and Lodge’s notion. Furthermore, need for cognition (NFC) as the
general motivation to engage in thorough processing should foster a greater confirma-
tion bias accordingly, as proposed in the following.

Hypothesis 2 (H2): Information processing styles (a) CR and (b) NFC increase the
extent of selective exposure to attitude-consistent messages (confirmation bias).

Moreover, affect as a situational factor will be considered because Festinger (1957)
suggested that the confirmation bias results from a desire to reduce or avoid unpleas-
ant states of dissonance; indeed, earlier work (Jonas, Graupmann, & Frey, 2006) dem-
onstrated that a negative mood led to a stronger confirmation bias.

Hypothesis 3 (H3): A less positive affective state increases the extent of selective
exposure to attitude-consistent messages.

Another lens in studying selection biases in political information use has applied a
social identity perspective to the phenomenon. For example, source cues have often
served to lead media users to anticipate attitude-consistent versus attitude-discrepant
messages (e.g., Iyengar & Hahn, 2009)—Relatedly, source and media perceptions
have recently been interpreted from a social identity perspective because media users
tend to be more familiar with sources that are aligned with their political views and
may perceive them as ingroup sources (Stroud et al., 2014). This line of work utilized
partisanship as a type of social identity. To extend this approach, the next section will
explore the social identity perspective with regard to nationality.

Ingroup Bias: A Social Identity Perspective

Numerous scholars have drawn upon social identity theory to study news content (Jones
& Sheets, 2009; Rivenburgh, 2000) and consumption (e.g., Gilmore, Meeks, & Domke,

4 Communication Research 00(0)

2013; Hoffner & Rehkoff, 2011). Social identity theory suggests that individuals cate-
gorize themselves and others into groups, with those groups they consider themselves
to be a part of as ingroups and the others as outgroups (Tajfel & Turner, 1979). Hence,
individuals think of themselves as having not only an individual “personal identity” but
also a “social identity” defined by these group memberships. The theory further postu-
lates ingroup favoritism such that individuals aim to derive positive self-esteem from
differentiating their ingroups from relevant comparison outgroups such that the com-
parison sheds a favorable light on the ingroup. Several studies have fruitfully applied
this perspective to news exposure. For example, Hoffner and Rehkoff (2011) inter-
preted voters’ perceptions of preelection coverage and found that Republicans per-
ceived the outgroup, defined as partisans of a nonpreferred party, to be more influenced
by the news coverage. Work by Appiah, Knobloch-Westerwick, and Alter (2013) dem-
onstrated that African American news users spent more time with news that presented
African Americans positively and news that presented Whites negatively, in line with
the ingroup favoritism hypothesis. Moreover, Knobloch-Westerwick and Hastall (2010)
found that Germans from an older age group preferred reading negative news about
younger individuals and positive news about same-age individuals, again corroborating
ingroup favoritism. The authors suggested that this bias might be particularly strong
among members of a group who perceive the comparison group to have higher status.

Nationality as a group identity should be highly relevant for news coverage and
consumption, given that much public and international affairs coverage pertains to
international relationships and comparisons. For example, Rivenburgh (2000) found
that news media tend to present the nation of their readership in a favorable light,
which would stand to reason in light of social identity theory, if readers prefer news
content that reflects positively on their social group defined by nationality. In light of
these theoretical considerations and some evidence on an ingroup bias in news selec-
tion, we extend this line of work to examine how national identity affects selective
exposure to news.

Hypothesis 4a (H4a): Selective exposure is greater for political messages that
praise the message recipients’ country in comparison with another country over
messages that critique it (ingroup bias).

Moreover, the social identity perspective suggests that the ingroup bias should be
stronger among individuals that identify more strongly with their nation as a social
group. In addition, social identity theory has its foundation in social comparison the-
ory (as mentioned above), which implies that the social identity impact should be
greater among those who generally tend to contrast themselves with others as an indi-
vidual social comparison tendency (Gibbons & Buunk, 1999).

Hypothesis 5 (H5): Stronger national identity among individual recipients increases
their selective exposure to messages that praise the message recipients’ country.
Hypothesis 6 (H6): Greater tendency to engage in contrasting social comparisons
increases selective exposure to messages that praise the message recipients’ country.

Knobloch-Westerwick et al. 5

Negativity Bias: A News Value Perspective

Interestingly, the above outlined social identity perspective, which implies that news
consumers favor positive news about their own group, challenges well-established
conceptualizations of what makes an event relevant for news coverage and consump-
tion. News values (Shoemaker, 1996) such as negativity of the portrayed events are
thought to not only affect what events are chosen to be covered in the news but to also
shape individuals’ selective news exposure (Eilders, 2006). Oftentimes, an evolution-
ary perspective is used to explain this overarching effect (Shoemaker, 1996; Soroka &
McAdams, 2015), arguing that humans are hardwired to watch out for threats in the
interest of survival and respond more strongly to negative stimuli with greater atten-
tion and stronger emotional responses.

Hence, negative news is generally thought to garner more attraction, which has
been supported by selective exposure studies on general news (e.g., Zillmann, Chen,
Knobloch, & Callison, 2004; Zillmann, Knobloch, & Yu, 2001). But also for political
news, the negativity concept has garnered much attention (see review by Lengauer,
Esser, & Berganza, 2012). Furthermore, some selective exposure research has detected
a negativity bias in different political information contexts: In a study by Meffert,
Chung, Joiner, Waks, and Garst (2006), participants were presented with a mock elec-
tion campaign for which information was presented through an electronic information
board. The participants generally favored negative news items, while a confirmation
bias in selecting messages was less pronounced. A study by Donsbach (1991) yielded
a similar result: Based on a Starch test with German newspaper users, the author found
that readers preferred attitude-consistent over attitude-discrepant articles mainly if
those articles framed political messages in a positive way. For negative articles,
instead, the confirmation bias almost disappeared. Given that the negativity bias along
with this news values perspective appears to alleviate the ingroup bias as outlined in
H4a, a competing hypothesis on a negativity bias will be examined.

Hypothesis 4b (H4b): Selective exposure is greater for political messages that
critique the message recipients’ country in comparison with another country over
messages that praise it (negativity bias).

Method

Overview

Hypotheses were tested in a laboratory experiment with 156 student participants at a
large U.S. Midwestern university in 2015. Participants first answered an online question-
naire that measured their social identity and attitudes toward eight controversial political
target issues of gay marriage, minimum wage, universal healthcare, gun ownership,
immigration, social welfare, national defense spending, and abortion (alongside distrac-
tor topics), as well as other variables such as affect measures. Upon completion of the
questionnaire, respondents were asked to browse an online news magazine that dis-
played headlines and leads of eight articles, one for each political topic, on its overview

6 Communication Research 00(0)

page. The display order of article leads on the overview page was randomized to avoid
any order effects. The articles were manipulated in a 2 × 2 within-subjects design, U.S.
stance (pro US vs. contra US) × partisan stance (liberal vs. conservative), with two arti-
cles for each combination—one article per news topic. Thus, four articles conveyed a
positive image of the United States in comparison with another country (“pro US”), and
four articles cast the United States in a negative light as compared with another country
(“contra US”). Among the four “pro US” and the four “contra US” articles, each featured
an either liberal or conservative view on the topic. The articles thus varied in the extent
to which they praised or criticized the United States and whether the presented view was
in line with a liberal or conservative stance on the issue.

The news magazine was presented to participants for 5 minutes. Within the given
time, participants were free to navigate between the overview page, which displayed
article headlines and leads, and the linked full texts. While browsing the online news
magazine, participants’ selective exposure to each article was unobtrusively logged.
Following the browsing period, respondents completed a second questionnaire with
measures of NFC, CR, and social comparison tendency, among other variables.

Participants

American student participants were recruited through instructors of introductory com-
munication classes and a research participant pool at a Midwestern university in the
United States. Student participants received course credit or extra credit for their par-
ticipation. Upon sign-up, students were invited to schedule an appointment for the
laboratory session. This procedure led to an initial sample of 197 cases. Two partici-
pants took the study twice, so their second entries were removed from the dataset.
Three cases with incomplete data entries due to technical errors were removed as well,
for a final sample of 192 complete cases. Furthermore, data were screened for partici-
pants who indicated that they have not lived in the United States since birth (n = 36),
for they may not perceive their U.S. identity as an equally important ingroup factor as
participants who were born in the United States.

The final sample thus consisted of 156 complete and valid cases. The mean age was
20.56 (SD = 1.60), and 62.2% of participants were female. The sample was 85.9%
White/Caucasian, 3.8% Asian, 3.2% African American, and 3.2% Hispanic or Latino;
further 2.6% of participants had a multiracial background. One percent indicated an
“other” ethnicity. Twelve percent of participants had a bachelor’s degree, 19.2% a
2-year college degree, and 69.2% some high school degree. With regard to their party
affiliation, 44.9% of participants indicated they feel closer to the Democratic party,
39.1% felt closer to the Republican party, and 16% were undecided.

Procedure

The study was administered in a laboratory session of about 30 minutes. A Microsoft
Silverlight application was specifically developed for the browsing task. The applica-
tion displayed the article leads on the overview page of the online news magazine in

Knobloch-Westerwick et al. 7

randomized order, and recorded browsing activity for each participant. Pre- and post-
selective exposure questionnaires were set up with standard commercial online survey
software (Qualtrics).

Participants obtained a personalized respondent ID in order to merge their responses
with their browsing data after completion of the study. The first questionnaire solicited
measures of participants’ social identity, political attitudes, partisanship, and affect,
among others. After completing the first questionnaire, participants were automatically
redirected to the Microsoft Silverlight application that administered the news browsing
task and displayed the news magazine overview page with article leads about eight dif-
ferent topics. Participants were asked to avoid any distractions and freely choose to read
whatever they are interested in. They were further briefed that they will only have a
limited time to browse the online news magazine, but not informed about the specific
amount of time. The study allotted 5 minutes for news browsing. Once browsing was
concluded, participants answered a second questionnaire measuring NFC, CR, social
comparison orientation, as well as demographics and other variables.

Stimuli and Pretest

The news browsing stimuli were specifically developed and pretested for the present
study. Stimuli pertained to eight different political topics and were shown in the con-
text of an online news magazine (for an illustration, see Figure 1). The eight article
leads, which appeared on the news magazine overview page, and their related full
texts varied in a 2 × 2 within-subjects design, U.S. stance (pro US vs. contra US) ×
partisan stance (liberal vs. conservative), with two articles per combination. Four arti-
cle leads praised the United States in contrast to another country (e.g., “America’s
National Defense and Spending Is Far Superior to UK’s”); the other four criticized the
United States while praising another country’s policy (e.g., “France Has Universal,
Efficient Healthcare and Outperforms America”). Two of the four “pro US” and “con-
tra US” leads conveyed a liberal view on the issue, whereas the other two “pro US”
and “contra US” leads took a conservative stance on the given issue.

The headlines and leads of each of the eight articles were presented on the overview
page of the online news magazine. The display order of news leads on the overview
page was randomized in order to avoid any confound between article content and article
position. Participants could click on the leads to get to the respective full articles. From
there, they could go back to the overview page at any time by clicking a return button.

The news articles that participants could choose to read were taken from news and
opinion webpages and edited for length, U.S. support/criticism, and partisan stance. All
articles were composed of a headline (Mwords = 9.13, SD = 0.99), a short article lead with
an average of 27.38 words (SD = 1.06), and a body of M = 799.25 words (SD = 1.75).

All news leads were pretested in a preceding survey with 15 American undergradu-
ate students (Mage = 21.4, SD = 0.51; 20% male). The pretest results confirmed that the
manipulation of article leads worked as intended. The average ratings for the item
“Does the article with this lead criticize or praise America?” (1 = criticize America; 11
= praise America) were between M = 8.60 and M = 9.47 for the “pro US” leads and

8 Communication Research 00(0)

between M = 2.53 and M = 3.53 for the leads of the “contra US” articles; details are
reported in the first column of Table 1. As a within-group factor in a repeated-mea-
sures ANOVA, the “pro US” versus “contra US” differentiation was significant, F(1,
14) = 88.93, p < .001, η2 = .864.

The average ratings for the item “Does the article with this lead align with the view
of the Democratic party or the Republican party?” (1 = Democratic party; 11 =
Republican party) were between M = 4.20 and M = 5.80 for the “liberal” leads and
between M = 6.87 and M = 7.80 for the “conservative” leads. A repeated-measures
ANOVA with partisan stance as a within-group factor revealed significant differences
between “liberal” and “conservative” leads, F(1, 14) = 9.53, p = .008, η2 = .405. Thus,
the manipulation of the leads’ partisan stance also worked as intended (for further
details, see second column of Table 1).

The eight news leads did not significantly differ on the item “How interesting is this
article lead to you personally?” (1 = not interesting at all; 11 = extremely interesting),
F(1, 14) = 0.100, p = .756. “Liberal” leads and “conservative” leads were perceived as
being equally interesting, F(1, 14) = 2.055, p = .174. The “pro US” and “contra US”
leads did not significantly differ in their level of being interesting, either, F(1, 14) =
1.280, p = .277; details are reported in the third column of Table 1.

Measures

The appendix presents a correlation matrix, means, standard deviations, and reliability
information of the measures.

National identity. To capture participants’ perceived relevance of being American, par-
ticipants rated various social identity characteristics with regard to their importance to

Figure 1. Example screenshot of the online magazine overview page.

Knobloch-Westerwick et al. 9

participants’ sense of who they are. The items were adopted from Cheek, Tropp, Chen,
and Underwood (1994), adapted for the purpose of the present study, and measured on
single-item scales ranging from 1 = not at all important to 9 = very important. The two
items “My feeling of pride in my country” and “Being an American citizen” served as
indicators of participants’ national identity. Both items were highly correlated (r = .67,
p < .001) and averaged for a preliminary national identity score (M = 6.93, SD = 1.93). Eight further items described additional group identity components (e.g., “My reli- gion” or “My sexual orientation, whether heterosexual, homosexual, or bisexual”). These items were averaged to obtain a social group identity score (Cronbach’s α = .87) that goes beyond participants’ national identity (M = 6.99, SD = 1.41). To capture the relative importance of participants’ national identity for their reading behaviors, in comparison with other identity facets, the group identity index was subtracted from the national identity score (M = −0.07, SD = 1.40).

Political attitudes. Participants’ political attitudes on the eight target issues served to
categorize selective exposure to each presented article as attitude-consistent or atti-
tude-inconsistent. Before the browsing task, participants were asked to indicate their
opinion on eight target issues, alongside four distractor topics, on one-item scales
ranging from 1 = strongly oppose to 8 = strongly support. For each topic, participants
answered the question “How strongly do you oppose or support the following poli-
cies?” Descriptive statistics for target attitudes are reported in the appendix.

Table 1. Stimulus Pretest for Online News Article Leads.

Article headlines

Pro US Partisan stance

M SD M SD

Liberal Pro US
Compared to Canada, U.S. Takes Highroad Aiding

Young Undocumented Immigrants
8.6 2.0 4.2 2.4

Japan Can Learn Much From U.S. Minimum Wage 9.5 1.1 5.8 2.

0

Liberal Contra US
France Has Universal, Efficient Healthcare and

Outperforms America
2.7 2.0 4.9 2.6

U.S. Economy Desperately Needs Germany’s Strong
Welfare Approach

2.9 1.8 4.3 2.6

Conservative Pro US
America’s National Defense and Spending Is Far

Superior to UK’s
9.3 1.2 7.1 2.9

U.S. Grants Right to Self-Defense While Netherlands
Endanger Citizens

9.1 1.9 7.8 3.0

Conservative Contra US
Unlike the U.S., Australia Has It Right with Pro-Life

Policy
3.5 3.0 7.0 3.4

South Korea Ensures Moral Stability, Prohibits Gay
Marriage—U.S. Flounders

2.5 2.0 6.9 3.4

10 Communication Research 00(0)

Selective exposure. While participants browsed the online articles, the online software
tracked their choices and how long they spent on individual article pages. On average,
participants selected M = 2.91 (SD = 1.57) of the eight displayed articles. For each
presented article, the exposure to the article was categorized as attitude-consistent if an
individual had given an attitude rating in the range of 5 to 8 as indicating support of
the promoted policy. For example, the article about abortion policies with the headline
“Unlike the U.S., Australia Has It Right With Pro-Life Policy” was categorized as
attitude-consistent if a participant had answered the question “How strongly do you
oppose or support: Anti-abortion law?” with a response option of 5, 6, 7, or 8 on an
eight-point scale with the labels 1 = strongly oppose and 8 = strongly support. It is
important to note that this approach created varying numbers of attitude-consistent
messages for each person (in contrast to other studies that presented messages for both
stances of each issue, for example, Knobloch-Westerwick & Meng, 2009). Per this
definition, the participants had on average M = 3.69 (SD = 1.68) attitude-consistent
articles available for selection and further reading.

In addition to attitude consistency, the selective exposure times were differentiated
by U.S. support, such that four selective exposure measures were computed: (a) atti-
tude-consistent pro U.S. messages, (b) attitude-discrepant pro U.S. messages, (c) atti-
tude-consistent contra U.S. messages, and (d) attitude-discrepant contra U.S. messages.
Within these four categories, the number of selected articles was summed up (for
number of selected articles as dependent variable) or the reading times (for selective
exposure in seconds as dependent variable) were accumulated, respectively. It should
be noted that the number of available attitude-consistent articles varied based on indi-
viduals’ attitudes on the issues.

CR. Participants completed the CR test, adopted from Frederick (2005). The test con-
sisted of three questions that indicate the amount of conscious deliberation partici-
pants are willing to invest in cognitive elaboration. Each question is designed in a
way that implies a seemingly obvious, but wrong answer if participants rely on their
spontaneous thoughts, whereas the right response requires a more reflective approach
to solve the given problem (for example, “A bat and a ball cost $1.10 in total. The bat
costs $1.00 more than the ball. How much does the ball cost?”). For each respondent,
correct answers (e.g., “5 cents” for the above mentioned example) were coded as 1,
and wrong responses were coded as 0. The answers were summed into an additive
index, ranging from 0 = no correct responses to 3 = three correct responses (Cron-
bach’s α = .65). On average, participants answered 0.66 questions correctly (SD =
0.95). Only 6.4% of participants gave three correct answers, 14.7% answered two
questions correctly, 17.3% had one correct response, and a majority of 61.5% scored
0 on the CR index.

NFC. In addition to participants’ actual cognitive processing styles as reflected in their
CR scores, participants indicated their general interest in cognitive elaboration by
answering a short NFC scale developed by Epstein, Pacini, Denes-Raj, and Heier

Knobloch-Westerwick et al. 11

(1996). Participants answered the following four items on 5-point scales (1 = com-
pletely false, 5 = completely true): “I don’t like to have to do a lot of thinking” (reverse
coded), “I try to avoid situations that require thinking in depth about something”
(reverse coded), “I prefer to do something that challenges my thinking abilities rather
than something that requires little thought,” and “I prefer complex to simple prob-
lems.” The four items were averaged to create an overall index of NFC (M = 3.38, SD
= 0.67) that reached an acceptable reliability of Cronbach’s α = .70.

Affect. To capture participants’ affective state before browsing the news articles, partici-
pants were asked to indicate how they are feeling right now with regard to four positive
emotions (i.e., “proud,” “strong,” “confident,” and “happy”) and four items describing
negative feelings, such as “tense” or “uncomfortable.” Participants indicated their
affect on 7-point Likert-type scales (1 = does not apply at all; 7 = applies very much).
The four positive items formed a reliable scale (Cronbach’s α = .87) and were averaged
for an overall index of positive affect (M = 4.60, SD = 1.22). (The negative affect items
also formed a reliable index (Cronbach’s α = .85, M = 2.74, SD = 1.33) that was nega-
tively correlated with positive affect (r = −.31, p < .001). Analyses reported below uti- lize only the positive affect index; using the negative affect index largely mirrored their findings.)

Partisanship and partisanship strength. To measure partisanship, participants answered the
question “Which party do you feel closer to?” The scale for this item ranged from 1 =
Democrats to 7 = Republicans. On average, participants showed a neutral political stance
with a slight preference for the Democratic party (M = 3.78, SD = 1.80). In addition, a
measure of partisanship strength was derived from this response, with the most extreme
response options 1 and 7 recoded to 4 as the highest score for partisanship strength, then
2 and 6 recoded to 3, 3 and 5 recoded to 2, and 4 recoded to 1 (M = 2.54, SD = .95).

Social comparison tendency. As general social comparison orientations may affect
selective exposure to the news articles presented in this study, participants indicated
their social comparison tendency with regard to eight items that were adopted from
Gibbons and Buunk (1999). Social comparison orientation consists of two distinct fac-
tors, one of which reflects the tendency to find similarities with others (e.g., “I often
try to find out what others think who face similar problems as I face.”), whereas the
second factor represents the tendency to contrast one’s own attitudes and behaviors
with others (e.g., “If I want to find out how well I have done something, I compare
what I have done with how others have done.”). Participants indicated their agreement
with each statement on a Likert-type scale ranging from 1 (strongly disagree) to 5
(strongly agree). The four items included to represent the first dimension formed a
highly reliable score (Cronbach’s α = .75) and were averaged to create an overall index
of “affiliation” orientation (M = 3.97, SD = 0.61). The four items on the second dimen-
sion also reached high reliability (Cronbach’s α = .79) and were averaged for an index
of “contrast” orientation (M = 3.57, SD = 0.78).

12 Communication Research 00(0)

Results

Impacts of Attitude Consistency and U.S. Stance on Selective Exposure

Two ANOVAs, each with four repeated selective exposure measures regarding (a)
attitude-consistent pro U.S. messages, (b) attitude-discrepant pro U.S. messages, (c)
attitude-consistent contra U.S. messages, and (d) attitude-discrepant contra U.S. mes-
sages, were conducted (see means illustrated in Figure 2) while controlling for the
number of attitude-consistent articles that were available on an individual basis. The
first ANOVA used number of selected articles in each of the four selective exposure
categories, whereas the second utilized the selective exposure times.

Based on the number of selected articles, the first ANOVA showed a strong confir-
mation bias in line with H1, F(1, 154) = 103.06, p < .001, η2 = .401. The simple com- parison between average number of selected attitude-consistent articles (M = 1.47, SD = 1.17) with the average number of selected attitude-discrepant articles (M = 1.45, SD = 1.23) does not illustrate this effect well, because it was strongly qualified by how many attitude-consistent articles were available on an individual basis, F(1, 154) = 116.13, p < .001, η2 = .430, as the more were available the more were selected, r = .47, p < .001. Furthermore, in line with H4b and in contrast to H4a, participants selected articles that praised other countries (M = 1.69, SD = 1.03) more frequently than they selected articles that praised the United States (M = 1.23, SD = 1.03), F(1, 154) = 6.05, p = .015, η2 = .038. No other impacts approached significance in this analysis; no interaction between confirmation bias and negativity bias emerged.

Figure 2. Selective exposure to political news articles as a function of attitude-consistency
and negativity of news messages.

Knobloch-Westerwick et al. 13

Using selective exposure times, the second ANOVA again yielded a strong confir-
mation bias, F(1, 154) = 48.96, p < .001, η2 = .241, in line with H1. On average, par- ticipants spent M = 144 s (SD = 101) on attitude-consistent messages compared with M = 115 s (SD = 100) for attitude-discrepant messages. The extent of the confirmation bias depended on how many attitude-consistent articles were available, F(1, 154) = 45.21, p < .001, η2 =.227, as the confirmation bias was greater the more attitude-con- sistent messages were available for perusal (r = .48, p < .001). The stance (pro or contra US) again yielded a significant impact, F(1, 154) = 7.41, p = .007, η2 = .046, because in line with H4b and in contrast to H4a, participants spent their reading time primarily with messages critiquing the United States, namely M = 157 s (SD = 92) compared with M = 101 s (SD = 88) for messages praising the United States. No other impacts approached significance in this analysis; no interaction between confirmation bias and negativity bias emerged.1

Impacts on the Extent of the Confirmation Bias

Regression analyses examined selective exposure to attitude-consistent messages,
with the predictors CR and NFC per H2 and positive affect as per H3. Furthermore,
national identity (per H5) and social comparison tendency (per H6) were included in
the regression models. In addition, biological sex and number of available attitude-
consistent articles served as control variables, along with (per reviewer request) parti-
sanship and partisanship strength.

This regression model was used for both number of selected articles (per reviewer
request), as reported in the upper half of Table 2, and for selective exposure in seconds,
reported in the lower half of Table 2. Because the number of selected articles did not
appear to be sensitive enough as dependent measure, further explanation focuses on
analyses with selective exposure in seconds as dependent variable.

The overall regression model (with 10 residual df, 145 regression df, and 155 total
df) for selective exposure attitude-consistent messages in seconds was significant, p < .001, with R2 = .369 (details are reported in the first column of Table 2). H2a was sup- ported because greater CR led to a stronger confirmation bias, β = .18, p = .018. Furthermore, in line with H2b, greater NFC fostered a stronger confirmation bias as well, β = .14, p = .05. A more positive affective state reduced the confirmation bias, β = −.15, p = .034, in line with H3. Participants who tended to contrast themselves with others more showed a stronger confirmation bias, β = .20, p = .014. Last, partisanship affected the confirmation bias, β = .15, p = .045, as a greater leaning toward the Republican party fostered the confirmation bias, and naturally the number of attitude- consistent messages that were available did as well, β = .57, p < .001. No additional significant effects emerged from the analysis.

When the same regression model was used for exposure to attitude-discrepant mes-
sages, the results mirrored these findings, just with reversed direction of the impacts
(see second column of Table 2). Given that the support for H2b was just at cutoff for
significance, it is relevant to note that greater NFC led to significantly shorter expo-
sure to attitude-discrepant messages, β = −.18, p = .017, which further supports H2b.

14 Communication Research 00(0)

Impacts on the Extent of Ingroup Versus Negativity Bias

Additional regression analyses investigated the exposure to messages praising the
United States, with national identity (H5) and social comparison tendency (H6) as
predictors while controlling for partisanship and partisanship strength. Furthermore,
CR, NFC, and positive affect were included in the model, as well as biological sex and
number of available attitude-consistent articles, as control variables. Again, while
regression analyses with articles selected as criterion variable were conducted and are
reported in the upper half of Table 2, further explanations focus on the analyses with
selective exposure in seconds as more sensitive criterion variable.

The overall regression model (with 10 residual df, 145 regression df, and 155 total
df) with selective exposure to articles praising the United States in seconds was signifi-
cant, p = .029, with R2 = .126 (see third column of Table 2). In line with H6, individuals
with the tendency to engage more in contrasting social comparisons spent more time

Table 2. Impacts on Selective Exposure to Political News (N = 156).

Attitude-
consistent

Attitude-
discrepant Pro US Contra US

Number of articles selected
Cognitive reflection .06 (.416) −.15 (.052) −.08 (.352) −.03 (.763)
Need for cognition .11 (.171) −.01 (.897) .10 (.288) .01 (.889)
Positive affect −.09 (.224) .07 (.345) −.06 (.510) .04 (.639)
National identity −.10 (.194) −.05 (.555) −.01 (.888) −.15 (.075)
Contrast orientation .17 (.046) −.09 (.296) .12 (.220) −.04 (.708)
Affiliation orientation −.15 (.071) −.01 (.920) −.10 (.316) −.08 (.375)
Partisanship .17 (.032) −.21 (.009) .08 (.377) −.14 (.119)
Partisanship strength −.01 (.901) −.02 (.773) −.01 (.930) −.03 (.733)
Gender (0 = male) −.07 (.352) .16 (.039) −.09 (.326) .20 (.023)
Available attitude-consistent

articles
.54 (<.001) −.49 (<.001) −.02 (.828) .04 (.614)

R2 .297 .288 .037 .102
Seconds spent reading
Cognitive reflection .18 (.018) −.17 (.020) .04 (.646) −.03 (.711)
Need for cognition .14 (.052) −.18 (.017) .10 (.235) −.13 (.129)
Positive affect −.15 (.034) .18 (.014) −.08 (.337) .10 (.224)
National identity .03 (.631) −.04 (.562) .10 (.259) −.10 (.246)
Contrast orientation .20 (.014) −.15 (.057) .18 (.049) −.12 (.184)
Affiliation orientation −.08 (.323) .04 (.642) −.09 (.335) .04 (.672)
Partisanship .15 (.045) −.19 (.013) .12 (.186) −.15 (.093)
Partisanship strength −.01 (.863) .06 (.426) .04 (.605) .01 (.934)
Gender (0 = male) −.12 (.106) .10 (.151) −.25 (.004) .22 (.011)
Number of attitude-consistent

articles
.57 (<.001) −.56 (<.001) −.01 (.886) .03 (.713)

R2 .369 .361 .126 .124

Note. Beta weights with p values in parentheses.

Knobloch-Westerwick et al. 15

with messages praising the United States β = .18, p = .049. The only additional impact
emerged for biological sex, β = −.25, p = .004, as men spent more time with messages
praising the United States than women did. When the same regression model was used
for exposure to messages critical of the United States, the only effect resulted from
biological sex, β = .22, p = .011 (for further details, see fourth column of Table 2).

Discussion

The present work tested hypotheses regarding selective exposure to political online news
articles that postulated a confirmation bias, an ingroup bias, and a negativity bias. The
biases under examination were proposed to be influenced by various media user charac-
teristics and circumstances—information processing styles, affective states, national
identity, and social comparison tendencies. American participants selected from online
news reports with either liberal or conservative stance on eight different topics that either
praised or critiqued the United States. Attitude-consistent exposure was defined based
on topic-specific attitudes reported before the online browsing task. Because each topic
was featured in only one news report with a particular stance, the number of attitude-
consistent messages that were available varied across participants (which is a method-
ological difference compared with other studies that defined attitude-consistent exposure
based on tracked selection of specific messages, as prior studies featured several mes-
sages on each topic with opposing stances; for example, Knobloch-Westerwick & Meng,
2009; Westerwick, Kleinman, & Knobloch-Westerwick, 2013).

The results yielded a strong confirmation bias in line with H1, which was greater if
an individual had more attitude-consistent messages to choose from. Less positive
affect also led participants in engaging in longer attitude-consistent exposure (support-
ing H3). This finding complies with earlier research (Jonas et al., 2006) and corrobo-
rates Festinger’s (1957) assumption that individuals entertain confirmation biases
mainly to enhance affective states or to avoid negative mood, as attitude-discrepant
messages may amplify cognitive dissonance and thus psychological discomfort.

More recently, bolstering one’s social identity defined per partisanship has also
been suggested as an explanation for the confirmation bias (e.g., Knobloch-Westerwick
& Meng, 2011; Stroud et al., 2014). Compatible with this view, participants who
tended to contrast themselves with others more through social comparisons showed a
stronger confirmation bias—They may feel more challenged by the notion that another
group may be right about a political issue.

In line with H2, both greater CR (per Frederick’s, 2005, CR test) and greater NFC
produced a stronger confirmation bias. This pattern corroborates the notion that “Being
a motivated reasoner takes effort” (Taber & Lodge, 2006, p. 757) because individuals
with the tendency to overcome intuitive but incorrect heuristics per CR engaged in more
attitude-confirming exposure, and so did individuals who tend to enjoy engaging in com-
plex thinking.

The support regarding H2—that those with greater thinking ability and motivation
exhibit a greater confirmation bias—might be particularly disturbing. If advanced
thinkers tend to fall victim to the confirmation bias when selecting political informa-
tion, chances of balanced consideration and contemplation on political issues in a

16 Communication Research 00(0)

democracy appear slim. It is possible, however, that the interaction with information
processing styles further depends on how much cognitive effort is required in a given
situation. In the present study, participants were presented with messages on eight dif-
ferent controversial topics; thus, relatively high complexity was presented. In a sim-
pler context, individuals with high CR or high NFC might be more open to engaging
with discrepant information, which might be more easily refuted when the stimulus
environment is less involving. Furthermore, a significant negativity bias in line with
H4b materialized, because participants generally preferred messages that critiqued
their country, which contradicts the ingroup bias expected per H4a. The only indica-
tion that influences per social identity theory and intergroup competition might affect
selective exposure to news emerged in form of an impact of individuals’ tendency to
contrast themselves more with others through social comparisons (in line with H6),
which led to longer exposure to news reports that praised the United States.

But overall, the general preference for U.S.-critical messages contradicts the social
identity framework—It appears that the ingroup bias was overridden by an interest to
watch out for problems for one’s own group along the lines of the negativity bias.
Moreover, the U.S.-critical messages may seem to carry greater informational utility
(e.g., Knobloch-Westerwick, 2008) along the lines of how one’s own country could
improve regarding the discussed political issue. Yet another consideration is that, per
news values framework (Shoemaker, 1996), U.S.-critical messages may have greater
novelty, as content analyses suggest that U.S. news coverage tends to present the
United States in a favorable light (Rivenburgh, 2000). Furthermore, Taber and Lodge’s
(2006) notion of a disconfirmation bias might suggest that individuals may spend
more time with a message that challenges their views because they are investing cog-
nitive effort to discount the message; yet the consideration that participants spent more
time with ingroup-critical messages to discount them does not match up with our find-
ings, because we did not find longer selective exposure times for outright counteratti-
tudinal messages (as H1 was confirmed), even when focusing the analysis on reading
times for articles an individual had selected in the first place.

When considering earlier empirical findings that supported a social identity per-
spective on an ingroup bias in selective news exposure, it becomes clear that low-sta-
tus groups in particular exhibited such a bias (Appiah et al., 2013; Knobloch-Westerwick
& Hastall, 2010)—possibly, the other featured countries did not seem like a suffi-
ciently relevant comparison group to the American participants or did not represent a
higher status group to them that might have triggered an ingroup as in these earlier
studies. When it comes to national political topics without reference to other nations,
political parties and partisanship should be a very salient ingroup/outgroup cue, which
was not present in the current stimuli/investigation. Thus, the social identity perspec-
tive should not be discounted based on our findings because partisan identity could
still be of major influence, and the salience of different identity facets is likely to vary
based on stimuli present. It is possible that the ingroup bias per national identity did
not emerge because of American exceptionalism beliefs (e.g., Lipset, 1996). Further
research is needed to test this assumption in an international context by including
nations with a less pronounced understanding of their nationality as an exceptional
ingroup factor.

17

A
p

p
e
n

d
ix

C
o
rr

el
at

io
n

M
at

ri
x,

M
ea

ns
, S

ta
nd

ar
d

D
ev

ia
ti

o
ns

, a
nd

C
ro

nb
ac

h’
s

A
lp

ha
s.

1
2

3
4

5
6

7
8

9
10

11
12

13
14

15
16

17
18

19
20

21
22

23
24

25
26

M
SD

1
.

A
C

SE
A

S
–

1.
43

1.
12

2
.

A
C

SE
T

S
.5

8*
**

–
14

7.
38

10
3.

44

3
.

A
D

SE
A

S
−

.2
4*

*
−

.6
6*

*

*
–

1.
39

1.
20

4
.

A
D
SE
T
S
−

.6
1*

**
−

.9
8*

*

*
.6

4*
**

–
11

8.
10

10
2.

05

5
.

Pr
o
U

S
SE

A
S

.4
8*

**
−

.0
3

.4
2*

**
−

.0
2

–
1.

17
0.

96

6
.

Pr
o
U
S
SE

T
S

.1
2

.1
4

−
.1

4
−

.1
4

.5
3*

**
–

10
2.

28
90

.0
1

7
.

Pr
o
A

B
S

E

A

S
.3

8*
**
−
.1

1
.5

4*
**

.1
1

.0
7

−
.5

4*
**
–
1.

66
1.

00

8
.

Pr
o
A
B
S

E
T

S
−
.1
4
−
.1

0
.1

0
.1

5
−

.5
8*

**
−
.9
8*

**
.5

3*
**

–
16

3.
21

92
.5

0

9
.

Pa
rt

is
an

sh
ip

.0
1

.0
0

−
.0

9
.0

0
.1

2
.1

6
−

.2
1*

−
.1

6
–

3.

78

1.
80

10
.

Pa
rt
is
an
sh
ip

s
tr

en
gt

h
−

.0
1
−
.1

0
.0

5
.1

0
.0

2
−

.0
3

.0
4

.0
4
−
.1

2
–

2.
54

0.

95

11
.

Ill
eg

al
im

m
ig

ra

ti
o
n

.1
7*

.1
3

−
.1
5
−
.1
4
−
.1
2
−

.1
6

.1
3

.1
5

−
.5
3*
**

.1
8*

–
4.

53
2.

08

12
.

M
in

im
um

w
ag

e
.1

9

*
.2

9*
**

−
.3

0*
**

−
.2

8*
*

−
.1

9*
−

.1
5
.0
4
.1
7*
−
.3
0*
**
.0
7

.3
0*

**
–

5.
59

2.
02

13
.

G
un

o
w

ne
rs

hi
p

.0
5

.2
0*

−
.2

1*
−

.1
9*

.0
6

.2
1*
−
.2

5*
*

−
.1

9*
.5

0*
**
.0
2
−
.2
8*
*
−
.0
6
–

4.
87

2.
61

14
.

A
nt

ia
bo

rt
io

n
.1

7*
.3

7*
**

−
.3
8*
**
−
.3

6*
**

−
.0
9
.0

0
−

.1
7*
.0
2

.5
1*

**
−

.1
0

−
.2

0

*
.0

5
.3

6*
**

–
3.

78
2.

62

15
.

S

o
ci

al
w

el
fa

re
.1

8

*
.1

3
−

.1
4
−
.1
3
−
.1
5
−
.2

1*
.1

9*
.1

9*
−

.6
0*

**
.1

5
.6

0*
**

.3
5*

**
−
.4
2*
**
−
.2
4*
*
–

4.

40

1.
91

16
.

G
ay

m
ar

ri
ag

e
−

.1
7*
−
.2

2*
.2

7*
*

.2
3*

*
−
.0
3
−
.1

4
.1

7*
.1

5
−

.5
5*

**
.1

0
.3

6*
**
.0
5
−
.3

5*
**

−
.5
5*
**

.2
5*

*
–

6.
73

2.
46

17
.

N
at

io
na

l h
ea

lt
hc

ar
e

.0
9

.0
7
−
.1

1
−

.0
6
−
.2

3*
*

−
.3
0*
**
.1
9*
.3
0*
**
−
.6
6*
**
.1

4
.5

6*
**

.4
1*

**
−
.5
3*
**
−

.2
8*

*
.6

2*
**

.3
3*

**
–

5.
07

2.
30

18
.

D
ef

en
se

s
pe

nd
in

g
−

.0
2
.0
6
−
.0
5
−
.0
6
.0
6
.0
0
−
.1

3
.0

0
.3
9*
**
−
.0
3
−

.2
2*

*
.0

3
.3

0*
**
.3
0*
**
−
.1
4
−
.2

0*
−

.2
6*

*
–

4.
32

1.
97

19
.

A
A

C
A

−
.4

8*
**
−
.4
9*
**

.4
7*

**
.4

8*
**
.0
9
.0
9
−
.0
6
−
.1
1
.2
2*
*
−
.0
5
−
.5

1*
**

−
.5
4*
**
−
.0
4
−
.2
2*
*
−

.4
6*

**
−
.1
5
−
.4
6*
**
−
.1

9*
–

3.
66

1.
68

20
.

Po
si

ti
ve

a
ffe

ct
−

.0
4
−
.0
3
−
.1
3
.0
2
−
.1
1
−
.0
5
−
.0
9
.0

4
.2

6*
*

.0
6
.0
2
−
.1
0
.1
3
.0

8
−

.0
6
−
.1
9*
−
.0
9
.2
3*
*
−
.0
7

.8
7

4.
60

1.
22

21
.

A
m

er
ic

an
id

en
ti
ty

−
.0

6
.0

2
−
.0
5
−
.0

2
.0

3
.0
8
−
.1
7*
−
.0

6
.2

9*
**
−
.1
0
−
.3
0*
**
−
.0
2
.1
5
.2
2*
*
−
.1
5
−
.1
5
−
.2
6*
*
.2
6*
*
.0
1
.1
4
.8
7

−
0.

07
1.

40

22
.

C
o
nt

ra
st

o
ri

en
ta

ti
o
n
.1
1
.1
5
−
.0
8
−
.1
2
.1
0
.1
3
−
.0

7
−

.1
0
.0
6

.0
8

.0
1
−
.1

4
.0

4
.0
1
−
.0
8
.0
3
−
.0

1
.0

3
−
.0
5
.0
8
−
.0

7
.7

9
3.

57
0.

78

23
.

A
ffi

lia

ti
o
n

o
ri
en
ta
ti
o
n
−
.0
2
.0
2
.0
1
−
.0
4
.0
1
.0
0
−
.0
2
−
.0
1
−
.0

6
.1

4
.1
1
−
.0
6
.0
2
−
.0

8
.0

7
.1

2
.0

5
.0

4
−
.0
5
.1
5
−
.1

6
.4

6*
**

.7
5

3.
97

0.
61

24
.

G
en

de
r

−
.0
1
−
.1
6
.2
3*
*
.1
5
.0
0
−
.2
8*
*
.2
7*
*
.2
6*
*
−
.0
8
.1
4
.1
0
.0
4
−
.2

2*
−

.0
6
.1
6
.2
5*
*
.0
5
.1
4
−
.0
9
−
.1
0
−
.0
3
.0
0
−
.0

4
–

1.
62

0.
49

25
.

C
R

.1
3
.2
5*
*
−
.1
9*
−
.2
5*
*
−
.0
5
.1
3
−
.0
3
−
.1
2
−
.1
1
−
.1
1
.0
6
−
.0
7
.0
6
−
.1
1
−
.0
2
−
.0
3
.0
1
−
.2
0*
.0
0
−
.0
3
−
.0
5
.0
1
−
.0
3
−
.3
3*

**
.6

5
0.

66
0.

95

26
.

N
FC

−
.0
1
.0
5
.0
1
−
.0
7
.0
2
.0
9
−
.0
1
−
.1
0
−
.0
3
.0
0
.0
6
−
.2
1*
.0
1
−
.1
0
−
.0
9
.0
8
−
.1
0
−
.1
1

.1
9

.1
0
−
.1

8*
−

.1
3
.1
0
−
.1
6
.2

2*
*

.7
0

3.
38

0.
67

N
ot

e.
R

ep
o
rt

ed
o

n
th

e
di

ag
o
na

l a
re

r
el

ia
bi

lit
ie

s.
A

C
SE

A
S

=
a

tt

it
ud

e-
co

ns
is

te
nt

s
el

ec
ti
ve

e
xp

o
su

re
a

rt
ic

le
s

el
ec

ti
o
n;

A
C

SE
T

S
=

a
tt

it
ud
e-
co
ns
is
te
nt
s
el
ec
ti
ve
e
xp
o
su

re
t

im
e

sp
en

t;

A
D
SE
A
S
=
a
tt
it
ud

e-
di

sc
re

pa
nt

s
el
ec
ti
ve
e
xp
o
su
re
a
rt
ic
le
s
el
ec
ti
o
n;

A
D

SE
T
S
=
a
tt
it
ud
e-
di
sc
re
pa
nt
s
el
ec
ti
ve
e
xp
o
su
re
t
im
e
sp
en

t;
P

ro
U

S
SE
A
S

=
p

ro
U

.S
. s

el
ec
ti
ve
e
xp
o
su

re

ar
ti
cl

e
se

le
ct

io
n;

P
ro

U
S

SE
T
S
=

p
ro

U
.S

. s
el

ec
ti
ve
e
xp
o
su
re
t
im
e
sp
en
t;
P

ro
A

B
S
E
A
S
=
p
ro

a
br

o
ad

s
el
ec
ti
ve
e
xp
o
su
re
a
rt
ic
le
s
el
ec
ti
o
n;
P
ro

A
B

S
E

T
S

=
p

ro
a

br
o
ad

s
el
ec
ti
ve

ex

po
su

re
t
im
e
sp
en

t;
I
lle

ga
l i

m
m

ig
ra

ti
o
n

=
s

up
po

rt
h

el
pi

ng
u

nd
o
cu

m
en

te
d

yo
un

g
im

m
ig

ra
nt

s;
M

in
im

um
w

ag
e

=
s
up
po

rt
a

m
in

im
um
w
ag

e;
G

un
o

w
ne

rs
hi

p
=

s
up

po
rt

g
un

o
w
ne
rs

hi
p;

A

nt
ia

bo
rt

io
n
=
s
up
po
rt
a
nt
ia
bo
rt
io
n

la
w

s;
S

o
ci
al
w
el
fa

re
=

s
up
po
rt

e
xt

en
di

ng
s

o
ci
al
w
el
fa

re
; G

ay
m

ar
ri

ag
e
=
s
up
po

rt
g

ay
m
ar
ri

ag
e;

N
at

io
na
l h
ea
lt
hc

ar
e

=
s
up
po

rt
u

ni
ve

rs
al

he

al
th

ca
re

; D
ef

en
se
s
pe
nd
in

g
=

s
up
po
rt

in
cr

ea
si

ng
n

at
io

na
l d

ef
en

se
s

pe
nd

in
g;

A
A

C
A

=
a

va
ila

bl
e

at
ti
tu

de
-c

o
ns

is
te

nt
a

rt
ic

le
s;

C
R

=
c

o
gn

it
iv

e
re

fle
ct

io
n;

N
FC

=
n

ee
d

fo
r

co
gn

it
io

n.
*p

< .0

5.
*

*p
<

.0
1.

*
**

p
<

.0
01

.

18 Communication Research 00(0)

Acknowledgments

The authors thank Travis Filicky, Erin Gottsacker, Melissa Kaminski, Park Lukich, Xirui Mao,
and Shan Xu for their help with the data collection, testing of the research application, and
stimuli preparation.

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 received no financial support for the research, authorship, and/or publication of this
article.

Note

1. A reviewer suggested to run additional analyses based just on the articles a participant
had selected, to examine whether the disconfirmation bias postulated by Taber and Lodge
(2006) would be evident in that attitude-discrepant articles would have longer reading
times. Thus, the same ANOVA model was run with selective exposure times derived only
from articles that a participant had selected. It yielded no significant effects.

References

Appiah, O., Knobloch-Westerwick, S., & Alter, S. (2013). Ingroup favoritism and outgroup
derogation: Effects of news valence, character race, and recipient race on selective news
reading. Journal of Communication, 63, 517-534. doi:10.1111/jcom.12032

Bennett, W. L., & Iyengar, S. (2008). A new era of minimal effects? The changing foundations
of political communication. Journal of Communication, 58, 707-731. doi:10.1111/j.1460-
2466.2008.00410.x

Cheek, J. M., Tropp, L. R., Chen, L. C., & Underwood, M. K. (1994, August). Identity orienta-
tions: Personal, social, and collective aspects of identity. Paper presented at the meeting of
the American Psychological Association. Los Angeles, CA.

D’Alessio, D., & Allen, M. (2000). Media bias in presidential elections: A meta-analysis.
Journal of Communication, 50, 133-156.

D’Alessio, D., & Allen, M. (2002). Selective exposure and dissonance after decisions.
Psychological Reports, 91, 527-532.

Donsbach, W. (1991). Exposure to political content in newspapers: The impact of cognitive
dissonance on readers’ selectivity. European Journal of Communication, 6, 155-186.
doi:10.1177/0267323191006002003

Donsbach, W. (2009). Cognitive dissonance theory—Roller coaster career: How communica-
tion research adapted the theory of cognitive dissonance. In T. Hartmann (Ed.), Media
choice: A theoretical and empirical overview (pp. 128-149). New York, NY: Routledge.

Eilders, C. (2006). News factors and news decisions: Theoretical and methodological advances
in Germany. Communications, 31, 5-24. doi:10.1515/COMMUN.2006.002

Epstein, S., Pacini, R., Denes-Raj, V., & Heier, H. (1996). Individual differences in intuitive-
experiential and analytical-rational thinking styles. Journal of Personality and Social
Psychology, 71, 390-405. doi:10.1037/0022-3514.71.2.390

Knobloch-Westerwick et al. 19

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140.
Festinger, L. (1957). A theory of cognitive dissonance. Palo Alto, CA: Stanford University Press.
Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic

Perspectives, 19(4), 25-42. doi:10.1257/089533005775196732
Garrett, R. K. (2009). Politically motivated reinforcement seeking: Reframing the selec-

tive exposure debate. Journal of Communication, 59, 676-699. doi:10.1111/j.1460-
2466.2009.01452.x

Gibbons, F. X., & Buunk, B. P. (1999). Individual differences in social comparison: Development
and validation of a scale of social comparison orientation. Journal of Personality and Social
Psychology, 76, 129-142. doi:10.1037/0022-3514.76.1.129

Gilmore, J., Meeks, L., & Domke, D. (2013). Why do (we think) they hate us: Anti-Americanism,
patriotic messages, and attributions of blame. International Journal of Communication, 7,
701-721.

Hart, W., Albarracín, D., Eagly, A. H., Brechan, I., Lindberg, M. J., & Merrill, L. (2009).
Feeling validated versus being correct: A meta-analysis of selective exposure to informa-
tion. Psychological Bulletin, 135, 555-588. doi:10.1037/a0015701

Hoffner, C., & Rehkoff, R. A. (2011). Young voters’ responses to the 2004 U.S. presidential
election: Social identity, perceived media influence, and behavioral outcomes. Journal of
Communication, 61, 732-757. doi:10.1111/j.1460-2466.2011.01565.x

Iyengar, S., & Hahn, K. S. (2009). Red media, blue media: Evidence of ideological selectivity in
media use. Journal of Communication, 59, 19-39. doi:10.1111/j.1460-2466.2008.01402.x

Iyengar, S., Sood, G., & Lelkes, Y. (2012). Affect, not ideology a social identity perspective on
polarization. Public Opinion Quarterly, 76, 405-431. doi:10.1093/poq/nfs038

Jonas, E., Graupmann, V., & Frey, D. (2006). The influence of mood on the search for support-
ing versus conflicting information: Dissonance reduction as a means of mood regulation?
Personality and Social Psychology Bulletin, 32, 3-15. doi:10.1177/0146167205276118

Jones, T. M., & Sheets, P. (2009). Torture in the eye of the beholder: Social identity, news coverage,
and Abu Ghraib. Political Communication, 26, 278-295. doi:10.1080/10584600903053460

Kahan, D. M. (2013). Ideology, motivated reasoning, and cognitive reflection. Judgment and
Decision Making, 8, 407-424. doi:10.2139/ssrn.2182588.

Knobloch-Westerwick, S. (2008). Informational utility. In W. Donsbach (Ed.), International
encyclopedia of communication (pp. 2273-2276). Oxford, UK: Wiley-Blackwell.

Knobloch-Westerwick, S. (2012). Selective exposure and reinforcement of attitudes and
partisanship before a presidential election. Journal of Communication, 62, 628-642.
doi:10.1111/j.1460-2466.2012.01651.x.

Knobloch-Westerwick, S. (2015). The selective exposure self- and affect-management
(SESAM) model: Applications in the realms of race, politics, and health. Communication
Research, 42, 959-985. doi:10.1177/0093650214539173

Knobloch-Westerwick, S., & Hastall, M. R. (2010). Please your self: Social identity effects on
selective exposure to news about in- and out-groups. Journal of Communication, 60, 515-
535. doi:10.1111/j.1460-2466.2010.01495.x

Knobloch-Westerwick, S., & Meng, J. (2009). Looking the other way: Selective exposure to
attitude-consistent and counterattitudinal political information. Communication Research,
36, 426-448. doi:10.1177/0093650209333030

Knobloch-Westerwick, S., & Meng, J. (2011). Reinforcement of the political self through selec-
tive exposure to political messages. Journal of Communication, 61, 349-368. doi:10.1111/
j.1460-2466.2011.01543.x

20 Communication Research 00(0)

Lengauer, G., Esser, F., & Berganza, R. (2012). Negativity in political news: A review
of concepts, operationalizations and key findings. Journalism, 13, 179-202.
doi:10.1177/1464884911427800

Lipset, S. M. (1996). American exceptionalism: A double-edged sword. New York, NY: W.W.
Norton.

Meffert, M. F., Chung, S., Joiner, A. J., Waks, L., & Garst, J. (2006). The effects of negativity and
motivated information processing during a political campaign. Journal of Communication,
56, 27-51. doi:10.1111/j.1460-2466.2006.00003.x

Rivenburgh, N. K. (2000). Social identity theory and new portrayals of citizens involved in
international affairs. Media Psychology, 2, 303-329. doi:10.1207/S1532785XMEP0204_01

Shoemaker, P. J. (1996). Hardwired for news: Using biological and cultural evolu-
tion to explain the surveillance function. Journal of Communication, 46, 32-47.
doi:10.1111/j.1460-2466.1996.tb01487.x

Smith, S. M., Fabrigar, L. R., & Norris, M. E. (2008). Reflecting on six decades of selec-
tive exposure research: Progress, challenges, and opportunities. Social & Personality
Psychology Compass, 2, 464-493. doi:10.1111/j.1751-9004.2007.00060.x

Soroka, S., & McAdams, S. (2015). News, politics, and negativity. Political Communication,
32, 1-22. doi:10.1080/10584609.2014.881942

Stroud, N. J. (2010). Polarization and partisan selective exposure. Journal of Communication,
60, 556-576. doi: 10.1111/j.1460-2466.2010.01497.x

Stroud, N. J., Muddiman, A., & Lee, J. K. (2014). Seeing media as group members: An evalu-
ation of partisan bias perceptions. Journal of Communication, 64, 874-894. doi:10.1111/
jcom.12110

Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs.
American Journal of Political Science, 50, 755-769. doi:10.1111/j.1540-5907.2006.00214.x

Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin
& S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33-47). Monterey,
CA: Brooks/Cole.

Westerwick, A., Kleinman, S. B., & Knobloch-Westerwick, S. (2013). Turn a blind eye if you
care: Impacts of attitude consistency, importance, and credibility on seeking of politi-
cal information and implications for attitudes. Journal of Communication, 63, 432-453.
doi:10.1111/jcom.12028

Zillmann, D., Chen, L., Knobloch, S., & Callison, C. (2004). Effects of lead framing on
selective exposure to Internet news reports. Communication Research, 31, 58-81.
doi:10.1177/0093650203260201

Zillmann, D., Knobloch, S., & Yu, H. (2001). Effects of photographs on the selective reading of
news reports. Media Psychology, 3, 301-324. doi:10.1207/S1532785XMEP0304_01

Author Biographies

Silvia Knobloch-Westerwick (PhD, Hanover University of Music, Drama and Media,
Germany) is a professor at the School of Communication, The Ohio State University, and serves
as editor of COMMUNICATION RESEARCH. Her 2015 monograph, “Choice and Preference
in Media Use: Advances in Selective Exposure Theory and Research,” was published by
Routledge. She has written or coauthored over 100 peer-reviewed articles and book chapters on
media uses and effects in the contexts of news and political communication, science communi-
cation, health communication, entertainment, and new communication technologies.

Knobloch-Westerwick et al. 21

Cornelia Mothes is a visiting professor at the Department of Media and Communication, TU
Dresden, Germany. Before, she was a postdoctoral research fellow at the School of
Communication at The Ohio State University, funded by the German Academic Exchange
Service (DAAD). Her main research interests lie in the field of political communication, media
psychology, and journalism.

Nick Polavin is a PhD student at The Ohio State University. His primary research interests
concern the role of communication in decision making, particularly in the context of juries and
law.

Media Psychology, 19:

203

–223, 2016

Copyright © Taylor & Francis

ISSN: 1521-3269 print/1532-785X online

DOI: 10.1080/15213269.2014.1002941

The Role of Partisan Sources and Audiences’
Involvement in Bias Perceptions of

Controversial News

MIHEE KIM
Department of Journalism and Mass Communication, Sungkyunkwan University,

Seoul, South Korea

Based on the hostile media effect (HME), this 2 (audiences’ opin-

ion) � 2 (partisan source) � 2 (news valence) factorial experiment

(N D 229) investigated the effects of partisan sources and audi-
ences’ different types of involvement on bias perceptions of slanted

news coverage regarding a controversial issue. The results show

that participants rated a news article, regardless of its valence

(congruent vs. incongruent), as less biased when it was attributed

to a partisan source that was consistent with their own position

than when it came from a partisan source that was inconsistent

with their position. Moderating effects of value-relevant involve-

ment on the source effects were found. The effects of partisan

sources on bias perceptions were only significant among those

with moderate or high levels of value-relevant involvement. The

implications of the source effects and the role of value-relevant

involvement as a moderator of such effects were discussed.

Audiences’ perceptions of source have been known to be an important
factor in their judgments of a message (Hovland & Weiss, 1951). Audiences’
evaluations of news coverage also depend on which news outlet reports
the coverage (Sundar & Nass, 2001). Due to the revolution of information
technology, the number of available news outlets is dramatically increasing.
The growth of cable television and Internet news outlets has provided a
more fragmented media environment where audiences have more options
for news consumption (Iyengar & Hahn, 2009). News outlets with certain
partisan perspectives have emerged and competed for audiences’ attention
in this new environment. These partisan news outlets ‘‘fit the news within a

Address correspondence to Mihee Kim, Department of Journalism and Mass Communi-
cations, Sungkyunkwan University, 25-2, Sungkyunkwan-Ro, Jongno-Gu, Seoul, South Korea.
E-mail: mhkim2008@gmail.com

203

204 M. Kim

political narrative and create a coherent conservative or liberal interpretation
of the day’s events’’ (Levendusky, 2013, p. 566). In this process, partisan
sources, such as newspapers and television channels, which are known for
being conservative or liberal, create certain political agendas by providing
slanted news that is written in a biased way ( Jamieson, Hardy, & Romer,
2007). As a result, audiences of partisan sources often perceive the other
side to be flawed or duplicitous, which decreases support for bipartisanship
within the public (Levendusky, 2013).

With the rise of partisan media exposure, it is more important to under-
stand the role of partisan sources in audiences’ perceptions of news content,
especially in regard to their bias perceptions of news coverage of con-
troversial issues. Audiences’ bias perceptions of news coverage have been
examined within the framework of the hostile media effect (HME), which
refers to the tendency for those who are highly involved in a controversial
issue to perceive balanced news coverage of that issue as biased against their
opinions (Gunther, Christen, Liebhart, & Chia, 2001). The concept of the HME
has been extended to the concept of the relative HME, which focuses on bias
perceptions of slanted news stories (e.g., Gunther & Chia, 2001; Gunther &
Christen, 2002; Gunther et al., 2001). In a media landscape that contains
partisan news outlets, people have more opportunities to be exposed to
slanted news coverage of controversial issues. Therefore, examining how
people perceive slanted news stories is more important than it has ever been
before when exploring how controversial issues are communicated. The
relative HME provides persuasive explanations for bias perceptions of slanted
news stories. This suggests that the relative HME is increasingly relevant in
contemporary media environments. However, very little research has been
conducted regarding the relative HME. This study adds to the literature of the
relative HME by exploring the role of partisan sources in bias perceptions
of slanted news stories.

Moreover, this study extends past literature by examining certain mod-
erators of source effects on the relative HME. Audiences’ involvement in
a controversial issue is considered to be a required condition for the HME.
Prior research has examined the role of two types of audiences’ involvement,
value-relevant involvement and outcome-relevant involvement, in the HME
by using balanced news stories that do not contain any source information.
In order to extend the past literature, this study investigates the role of
audiences’ involvement types as moderators of the effects of source on the
relative HME. Most of the studies on the HME have utilized the concepts of
both group membership and attitude extremity as indicators of audiences’
involvement (Gunther, Miller, & Liebhart, 2009); however, these concepts do
not clearly differentiate the two different types of involvement (Choi, Yang, &
Chang, 2009). None of the past studies have consistently conceptualized and
operationalized audiences’ value-relevant and outcome-relevant involvement
in the context of the effects of source on the relative HME. In order to

Partisan Source and Involvement 205

eliminate this dearth of information, this study explores the role of audiences’
value-relevant and outcome-relevant involvement in the effects of partisan
sources (congruent vs. incongruent) on bias perceptions of slanted news
(congruent vs. incongruent).

THE HOSTILE MEDIA EFFECT (HME)

Balanced messages are considered to be biased by certain audiences. This
phenomenon has been explored within the framework of biased assimilation
or the HME. Lord, Ross, and Lepper (1979) documented that biased assimi-
lation provided evidence of those with opposing views perceiving research
results to be in favor of their points of view. In their study, participants
accepted congenial information but ignored uncongenial information that
was provided by a research report. This biased assimilation of information
resulted in the polarization, rather than the convergence, of opposing views.
Regarding partisans’ polarized worldviews within media coverage, Vallone,
Ross, and Lepper (1985) investigated how audiences with opposing positions
perceive balanced news coverage of a controversial issue. Unlike the results
of the biased assimilation studies, they found that partisan news consumers
perceive balanced news coverage of a controversial issue to be biased against
their own existing position, which is referred to as the HME. In the mass
media context, many empirical studies have vividly demonstrated the HME
across issues and populations (e.g., Christen, Kannaovakun, & Gunther,
2002; Giner-Sorolla & Chaiken, 1994; Gunther et al., 2001; Gunther et al.,
2009; Gunther & Liebhart, 2006; Gunther & Schmitt, 2004; Kim, 2010).

Additionally, the HME has been examined by using clearly slanted news
content. In an experimental study, Gunther and associates (2001) found that
when partisans are exposed to news that is unfavorable to their side, they
perceive the news to be more biased than partisans on the other side of
the issue do, which is referred to as the relative HME. Studies of the relative
HME (Gunther & Chia, 2001; Gunther & Christen, 2002; Gunther et al., 2001)
confirmed that ‘‘regardless of whether news content is objectively neutral,
opposing parties would have divergent perceptions of news slant’’ (Choi
et al., 2009, p. 56).

Previous studies of the relative HME (Gunther & Chia, 2001; Gunther &
Christen, 2002; Gunther et al., 2001) have demonstrated that audiences’ bias
perceptions of news content that contains a clear slant depend on whether
the news valence is congruent or incongruent with their own opinions. This
means that those who have opinions that are incongruent with the news
valence perceive the news story to be more biased than those who have
opinions that are congruent with the news valence. Therefore, the following
hypothesis is proposed in order to test the relative HME of news content
from partisan sources:

206 M. Kim

H1: Audiences who have opinions that are incongruent with the news va-
lence perceive news stories from partisan sources to be more biased
than audiences who have opinions that are congruent with the news
valence.

SOURCE EFFECTS WITHIN THE HME

Gunther and Schmitt (2004) explained the influence of sources on the HME
by using a reach concept. In their experiment, the HME was manifest when
content was attributed to a newspaper article that had a high perceived
reach, while biased assimilation was evident when the same content was
attributed to a student essay that had a low perceived reach. Gunther and
Schmitt’s study provided strong evidence that mass media source is a neces-
sary condition for the HME; however, the source concept in their study was
confounded with the reach concept.

In order to investigate this confound, Gunther and Liebhart (2006) at-
tempted to separate source (journalist vs. college student) from reach (mass
media vs. classroom composition) in their experimental design. They found
that both reach and source independently produced the HME. However, the
interaction between source and reach was found in manipulation checks,
which suggests possible confounding effects of the two factors on actual
audiences’ hostile media perceptions.

Unlike the above studies, which examined sources that had different
perceived reach (mass media vs. student essay), certain scholars compared
the effects of a mass media source with those of another mass media source,
controlling for perceived reach. For example, Arpan and Raney (2003) ex-
amined how different newspapers influence the HME in the context of
sports news and fanship. College students were exposed to a balanced story
about their hometown college football team in one of three newspapers: a
hometown paper, a cross-state rival university town’s paper, and a neutral-
town paper. The news from the rival town’s paper and the neutral paper
was perceived to be more biased than the same news from the home town
paper. In the same vein, Ariyanto, Hornsey, and Gallois (2007) exposed
Muslims and Christians to an article about inter-religious conflict. People
read the same article from three different newspapers: a Muslim newspaper,
a Christian newspaper, and an unidentified newspaper. While participants
perceived the article from the Muslim newspaper to be biased toward Mus-
lims, they regarded the same article from the Christian newspaper to be
biased toward Christians. These studies show that audiences’ prior beliefs
of a source influence their bias perceptions of news coverage from that
source.

A more recent study (Reid, 2012) tested the effects of source on the HME
in the context of partisan authors. Participants who were either Democrats

Partisan Source and Involvement 207

or Republicans read a balanced CNN news report on their candidates that
was written by either a member of a Democratic or Republican think tank.
When participants saw the news story that was written by authors who
had the same political position as their own, they believed that the news
story contained more favorable information about their own candidates. In
contrast, when they read the news story that was written by authors who
had different political views than their own, they believed that the news
story contained more favorable information about the other side’s candidates.
These findings suggest that while partisan authors who have similar political
positions to their audiences produce biased assimilation, partisan authors
who have different positions than their audiences generate the HME. In
other words, aspects of partisan authors, whose views are congruent or
incongruent with the audiences’ political positions, induce different bias
perceptions within the audiences regarding the same news coverage.

Further, Reid (2012) investigated how two partisan groups, Democrats
and Republicans, perceive an attack against Democrats, based on the views
of the authors of the attack. The attack was perceived to be less biased
when the author of the attack was a Democrat than when the author was
a Republican. By extending previous work that has shown the effects of
partisan authors of news content on the HME (e.g., Reid, 2012), this study
explores the effects of partisan sources, which are news outlets that have
certain partisan perspectives (congruent vs. incongruent), on audiences’ bias
perceptions of slanted news stories that are written in a biased way (con-
gruent vs. incongruent). Anchored in the body of literature that reveals the
effects of source on the HME, the following hypotheses are proposed:

H2a: Audiences perceive an incongruent news story as less biased when it
comes from a source that is consistent with their own position than
when it comes from a source that is inconsistent with their position.

H2b: Audiences perceive a congruent news story as less biased when it
comes from a source that is consistent with their own position than
when it comes from a source that is inconsistent with their position.

AUDIENCES’ INVOLVEMENT AND THE HME

A person’s involvement in an issue has been shown to be a predictor of
the HME. Two types of involvement, value-relevant and outcome-relevant
involvement, have been examined in HME studies. Value-relevant involve-
ment, which is often termed ego involvement, refers to ‘‘the psychological
state that is created by the activation of attitudes that are linked to important
values’’ ( Johnson & Eagly, 1989, p. 290). Outcome-relevant involvement is
conceptualized as ‘‘the relevance of an issue to one’s currently important
goals or outcomes’’ ( Johnson & Eagly, 1989, p. 292). While people with

208 M. Kim

value-relevant involvement in an issue form their own opinions based on
their social or personal values, people with outcome-relevant involvement
form their attitudes based on the future consequences of the issue (Choi,
Park, & Chang, 2011).

Social judgment theory provides a theoretical framework for under-
standing the relationship between audiences’ involvement types and the
HME. Social judgment theory posits that different levels of value-relevant
involvement with a given issue are related to different levels of acceptance,
rejection, or non-commitment regarding news coverage of an issue (Sherif,
Sherif, & Nebergall, 1965). According to the theory, those who are more ego
involved with an issue tend to have wider latitudes of rejection of statements
about the issue, whereas less ego-involved people have relatively narrow
latitudes of rejection but wider latitudes of acceptance or non-commitment.
While messages that are located within a reader’s latitude of acceptance
generate assimilation effects, messages that fall into the latitude of rejection
produce the HME (Sherif et al., 1965).

Unlike value-relevant involvement, increasing outcome-relevant involve-
ment does not raise the latitude of rejection ( Johnson, Lin, Symons, Camp-
bell, & Ekstein, 1995), which suggests that outcome-relevant involvement
activates different cognitive processes than value-relevant involvement does
(Cho & Boster, 2005; Johnson & Eagly, 1989). People with high outcome-
relevant involvement tend to adopt ‘‘a position that maximizes the immediate
situational rewards’’ (Petty & Cacioppo, 1979, p. 1916). Thus, these people
are known to be motivated to process messages systematically. Drawing
on this logic, Petty and Cacioppo (1979) empirically examined how those
with outcome-relevant involvement perceive disagreeable arguments. The
researchers found that increasing outcome-relevant involvement enhanced
the persuasion of counterattitudinal arguments when the arguments were
strong and compelling. This suggests that counterattitudinal messages may
not result in the HME, especially for individuals who have high outcome-
relevant involvement.

Therefore, it was expected that those with value-relevant involvement,
rather than outcome-relevant involvement, process messages in an opposi-
tional way, which results in the HME (Choi et al., 2011). Choi et al. (2009) pro-
vided empirical evidence for this prediction. In their study, value-relevant in-
volvement was a critical predictor of the HME, rather than outcome-relevant
involvement. The current study extends prior research by testing the effects
of these two involvement types in the context of the relative HME. Based
on social judgment theory and previous studies (e.g., Choi et al., 2009; Choi
et al., 2011), the following hypothesis is proposed:

H3: Value-relevant involvement has a greater effect on the perceived bias
of a slanted news story than outcome-relevant involvement.

Partisan Source and Involvement 209

AUDIENCES’ INVOLVEMENT TYPES AND THE

EFFECTS OF SOURCE ON THE HME

Few studies have examined the different influences of the two types of
involvement on the effects of source on the HME. Given the disparate
nature of value-relevant and outcome-relevant involvement, these two types
of involvement may have different effects on the role of partisan sources
in the HME. Thus, this study explores which type of involvement is more
sensitive to the effects of variations in source factors on bias perceptions of
slanted news.

Value-Relevant Involvement

Sherif and Sherif (1967) theorized the relationship between the source of
content and readers’ value-relevant involvement with a topic of the content.
The researchers argued that because people who are not highly ego involved
with an issue lack the internal standards to judge content, they use other
cues, such as the identity of the source. In other words, for those who are
less ego involved in an issue, the source of content may become a major
anchor for the evaluation of the content (Sherif & Sherif, 1967). Consistent
with this view, social judgment theory predicts that source effects will be
manifest among those with low levels of value-relevant involvement, rather
than those with high levels of value-relevant involvement (Sherif et al.,
1965).

Previous attitude change studies have provided evidence that supports
social judgment theory’s prediction. Sereno (1968) investigated the relation-
ship among value-relevant involvement, the high credibility of a source, and
attitudinal responses to a belief-discrepant message. When the message was
attributed to a highly credible source, participants with high value-relevant
involvement changed their attitude on the topic in the direction that was
advocated in the message less than participants with low value-relevant
involvement did. Along similar lines, Johnson and Scileppi (1969) found
a significant interaction between the source and value-relevant involvement.
There was greater attitude change in the low value-relevant involvement
and high source credibility condition than in the other three combinations
of source credibility and value-relevant involvement. The researchers ar-
gued that because participants with high value-relevant involvement evaluate
content more critically, source differences tend to disappear. Rhine and
Severance (1970) also found that more attitude change occurred for those
with low, rather than those with high, value-relevant involvement. For those
with high value-relevant involvement, source credibility had little effect on
attitude change.

Certain HME studies examined the relationship between audiences’ value-
relevant involvement and source effects. Gunther et al. (2009) investigated

210 M. Kim

the effects of source on the HME in the context of genetically modified wild
rice. Two groups (non-Native Americans vs. Native Americans) were exposed
to news from different sources (friendly vs. non-friendly newspapers). A
main effect of source on the perceived bias of the news was not found;
however, an interaction effect between source and group was significant.
Source effects were only manifest in the non-Native American group. Only
non-Native Americans judged the news to be more favorable to their posi-
tion when it came from a friendly source than when it came from a non-
friendly source. Native Americans perceived the same news to be neutral,
regardless of the source. The researchers argued that group membership
is conceptually related to audiences’ value-relevant involvement. If this is
true, the results suggest that those with high value-relevant involvement may
not be influenced by sources regarding their hostile media perception. The
effects of source on the HME may only be manifest among those with low-
levels of value-relevant involvement. This is consistent with the idea of social
judgment theory.

A more recent study by Reid (2012) explored how partisanship (Demo-
crats vs. Republicans) influences the effects of source on the HME. The effects
of source on the HME were amplified by partisanship, which, according to
the author, reflects audiences’ value-relevant involvement. This suggests that
when people have higher value-relevant involvement, the effects of source
on the HME will be greater. This is contrary to social judgment theory’s
prediction.

As shown above, previous studies provide conflicting evidence for the
relationship between value-relevant involvement and the effects of source
on the HME. On the one hand, when value-relevant involvement was op-
erationalized as group membership, the effects of source on the HME were
manifest among those with low levels of value-relevant involvement (Gun-
ther et al., 2009). On the other hand, when value-relevant involvement
was measured by partisanship, high levels of value-relevant involvement
increased the effects of source on the HME (Reid, 2012).

Partisanship and group membership do not explain what type of in-
volvement underlies audiences’ attitudes. A person may decide to belong to
one group because the group is related to one’s personal values or because
the group is currently related to one’s important goals or outcomes (Choi
et al., 2009). Partisanship can also be developed based on either value-
relevant or outcome-relevant involvement (Cho & Boster, 2005). Therefore,
it is unclear whether either partisanship or group membership distinctly in-
dicates audiences’ ego involvement with an issue. Accordingly, the influence
of different levels of ego-involvement (high vs. low) on the effects of source
on the HME needs to be investigated. This study attempts to measure distinct
constructs of audiences’ value-relevant involvement and then tests how this
involvement moderates the effects of source on bias perceptions of clearly
slanted news by examining the following research question:

Partisan Source and Involvement 211

RQ1: How does an audience’s value-relevant involvement in a news topic
moderate the effects of source on the relative HME?

Outcome-Relevant Involvement

Audiences’ outcome-relevant involvement was investigated in relation to
persuasive messages. Petty, Cacioppo, and Goldman (1981) found that for
those with high outcome-relevant involvement, the quality of arguments in
a message significantly influenced their attitudes. Attitudes of those with low
outcome-relevant involvement were primarily influenced by the expertise of
the source. This difference suggests that the effects of source on persua-
sive messages are greater among those with low outcome-relevant involve-
ment. In the same vein, Chaiken (1980) examined the relationship between
outcome-relevant involvement (high vs. low) and the information process
(heuristic vs. systematic). The results show that those with high outcome-
relevant involvement employed a systematic information processing strategy
in which message-based cognitions mediated persuasion. In contrast, partic-
ipants with low outcome-relevant involvement used a heuristic processing
strategy in which non-content cues, such as the source’s identity (likable vs.
unlikable), mediated persuasion. This difference in involvement suggests that
for those who employ the heuristic strategy, source characteristics may exert
a greater impact on the perceptions of messages than message character-
istics (Chaiken, 1980). These persuasion studies demonstrate that outcome-
relevant involvement moderates the effects of source on attitude changes.
This means that source effects may be maximized when the message is on a
topic that is not related to message receivers’ self-interest or important goals
(Petty & Cacioppo, 1979).

In regard to HME research, only Gunther et al.’s (2009) study examined
the role of outcome-relevant involvement in source effects, investigating
the relationship between attitude involvement and the effects of source on
the HME. In their study, as noted earlier, involvement that was based on
group membership interacted with source, but involvement that was based
on attitude measures did not interact with source. The researchers argued
that the attitude measures of their study shared elements with outcome-
relevant involvement. If this is true, outcome-relevant involvement may not
moderate the effects of source on the HME. This finding is not consistent with
the findings of persuasion studies. These conflicting results demonstrate the
need for further research that examines whether outcome-relevant involve-
ment moderates the effects of source on the HME. Therefore, the following
research question is proposed:

RQ2: Does an audience’s outcome-relevant involvement in a news topic
moderate the effects of source on the relative HME?

212 M. Kim

METHOD

This study’s focal issue was the dissolution of the Unified Progressive Party
in South Korea. The South Korean government accused the Unified Progres-
sive Party of participating in pro-North Korean activities and petitioned the
Constitutional Court to dissolve this group early in November 2013. Since this
occurrence, controversy (pro or con) over this issue has been accelerated in
South Korea. Supporters of the dissolution of the party welcomed the gov-
ernment’s move as a way to enhance national security, whereas opponents
claimed that the compulsory dissolution of the political party may take away
people’s political freedom.

This study employed an online experiment that had a 2 (audiences’
opinion: pro or con) � 2 (news valence: congruent or incongruent) � 2
(partisan source: congruent or incongruent) factorial design. In this experi-
ment, while audiences’ opinion variable was measured, both news valence
and partisan source variables were manipulated. This web-based experiment
was conducted during the last week of November 2013.

Participants

An online survey company in South Korea recruited participants. The survey
company sent email invitations to its 980,000 panel members. Initially, a total
of 483 participants signed up for this study. Because audiences who have
opinions that support or oppose the focal issue have been regarded to be a
necessary condition for the HME (Gunther et al., 2009), those who reported
holding a neutral position (N D 226) were excluded. In addition, data from
28 participants who did not complete this experiment were eliminated. This
selection produced a total sample of 229 participants (150 men, 79 women;
age: M D 42.17, SD D 10.63).

Procedure

Those who voluntarily wished to participate in this experiment were pro-
vided with a link to a website. They completed a pretest survey that asked
about their opinions, value-relevant involvement, and outcome-relevant in-
volvement in the focal issue. After that survey, they were randomly assigned
to one of four conditions, which included being given a news article (con-
gruent or incongruent) from a certain source (congruent or incongruent).
Random assignment to the conditions was automatically conducted by the
online survey tool. After reading the news article, they completed a post
survey that measured their perceived bias regarding the news article that
they had just read.

Partisan Source and Involvement 213

Stimuli

The valence (pro-dissolution of the Unified Progressive Party or anti-dissolu-
tion of the Unified Progressive Party) of the news articles was manipulated
based on the supporters’ or opponents’ arguments that were represented
in previously published news stories. The valence of the news articles was
previously tested by 18 college students. They all indicated that the news
articles were explicitly slanted as intended.

Two sources, Chosun Daily and Hankyoreh Newspaper, were selected
for this experiment. These two newspapers are considered to be partisan
news outlets that have opposite political ideologies in South Korea. While
Chosun Daily has a distinctive reputation for providing conservative or pro-
government perspectives, Hankyoreh Newspaper is widely known to deliver
more liberal or anti-government points of view. The news article page that
was attributed to Chosun Daily or Hankyoreh Newspaper was meant to
resemble an actual online news page from each newspaper.

Pretest Measures

Participants were asked to indicate their preexisting opinions on the issue
(i.e., ‘‘How strongly do you support or oppose the dissolution of the Unified
Progressive Party?’’) by using a scale from 1 (strongly oppose) to 7 (strongly
support). Two groups were finally identified: a supporter group (N D 117,
M D 6.13, SD D .87) and an opponent group (N D 112, M D 2.10, SD D
.83). There was a significant difference in opinions between the two groups,
t (227) D 36.66, p < .001.

Two types of involvement were measured by the pre-survey. Value-
relevant and outcome-relevant involvement are distinct attitude constructs
(Choi et al., 2009; Choi et al., 2011). This study conducted a confirmatory
factor analysis (CFA) in order to confirm the validity of the two types of
involvement constructs. As expected, the two-factor model was supported
(�2(4) D 8.29, p D .08, CFI D .994, RMSEA D .06). The one-factor model
did not fit the data (�2(5) D 257.26, p < .001, CFI D .656, RMSEA D .47).

In order to measure value-relevant involvement, participants were asked
to indicate how much they agreed with the following statements: ‘‘My posi-
tion on the dissolution of the Unified Progressive Party has a lot to do with
my beliefs about how life should be lived,’’ ‘‘My position on the dissolution
of the Unified Progressive Party reflects who I am,’’ ‘‘The values that are
most important to me are what determine my stand on the dissolution of the
Unified Progressive Party issue’’ (1 D strongly disagree, 7 D strongly agree).
The reliability was ˛ D .89 (M D 4.74, SD D 1.40).

Outcome-relevant involvement was measured by asking the extent to
which they agreed with the following statements: ‘‘Whether the Unified

214 M. Kim

Progressive Party is dissolved or not has little impact on my life,’’ ‘‘All in
all, the effect of the dissolution of the Unified Progressive Party on my life is
little,’’ (1 D strongly disagree, 7 D strongly agree). These two items provided
high reliability (˛ D .91, M D 4.10, SD D 1.57).

For source manipulation checks, participants were asked to evaluate two
newspapers that were employed as partisan sources in the subsequent exper-
iment, Chosun Daily and Hankyoreh Newspaper: ‘‘How would you describe
Chosun Daily’s (Hankyoreh Newspaper’s) political ideology?’’ (1 D strongly
conservative, 7 D strongly liberal), ‘‘How would you describe Chosun Daily’s
(Hankyoreh Newspaper’s) political position?’’ (1 D strongly against the gov-
ernment, 7 D strongly in favor of the government). In order to reduce
any priming effects, the items regarding this study’s target sources were
included with questions for other newspapers, such as Joongang Daily,
Donga Daily, and Kyunghyang Newspaper, which are leading newspapers
in South Korea.

Posttest Measures

In order to measure participants’ perceived bias, three items were adapted
from previous studies (e.g., Arpan & Raney, 2003). Participants were prompted
to indicate how much they agreed with the following statements using a 7-
point scale (1 D strongly disagree, 7 D strongly agree): ‘‘The portrayal of the
dissolution of the Unified Progressive Party in the news article is biased,’’
‘‘Chosun Daily (Hankyoreh Newspaper), which reported the news article, is
a biased news outlet,’’ ‘‘The reporter of the news article is biased.’’ All three
items, which produced high reliability (˛ D .91), were combined in order to
create a bias index.

As a manipulation check for content, participants were asked to report
their thoughts about the valence of the news article (i.e., ‘‘How strongly does
the news article support or oppose the dissolution of the Unified Progressive
Party?’’) by using a scale from 1 (strongly oppose) to 7 (strongly support).

RESULTS

Manipulation Checks

An independent t-test showed that the news article manipulated in favor of
the dissolution of the Unified Progressive Party was considered to support the
dissolution (M D 5.57, SD D 1.60). In contrast, the article that was designed
to be against the dissolution was perceived to oppose it (M D 3.46, SD D
2.07). The difference between the perceived valences of the news articles was
significant (t (227) D 8.62, p < .001). This check confirmed that participants perceived the valence of the news articles as intended.

Partisan Source and Involvement 215

A repeated-measures analysis of variance (ANOVA) was employed in
order to test the source manipulations. Chosun Daily was perceived to be a
conservative outlet (M D 2.05, SD D .08), while Hankyoreh Newspaper was
judged to be a liberal channel (M D 5.32, SD D .09). This difference was
significant (F (1,228) D 449.09, p < .001, �2p D .66). In addition, participants considered Chosun Daily to be in favor of the government (M D 5.73, SD D .10), while they regarded Hankyoreh Newspaper to be against the govern- ment (M D 2.95, SD D .09). This difference was also significant (F (1,228) D 352.15, p < .001, �2p D .61). As a whole, participants evaluated Chosun Daily to be a conservative and pro-government news outlet, whereas they considered Hankyoreh Newspaper to be a liberal and anti-government news source. This suggests that in the context of this study’s focal issue, Chosun Daily would be perceived to be in favor of the dissolution of the Unified Progressive Party, whereas Hankyoreh Newspaper would be considered to be against the dissolution.

Hypothesis Tests

Hypothesis 1 proposed that audiences with opinions that were incongruent
with the news valence would perceive news stories from the partisan sources
to be more biased than audiences with opinions that were congruent with the
news valence. The results of a one-way ANOVA demonstrated a significant
main effect of whether a participant’s opinion is consistent with the news
valence (F (1,227) D 29.39, p < .001, �2p D .12). As predicted in Hypothesis 1, those with an opinion that was incongruent with the news valence (M D 4.81, SD D .15) perceived the news to be more biased than those with a stance that was congruent with the news valence did (M D 3.69, SD D .14). Thus, Hypothesis 1 was supported.

Hypotheses 2a and 2b focused on the source effects on audiences’ bias
perceptions of clearly slanted news. Hypothesis 2a predicted that audiences
would perceive an incongruent news story as less biased when it came from a
source that was consistent with their own position than when it came from a
source that was inconsistent with their position. In support of Hypothesis 2a,
the one-way ANOVA confirmed a statistically significant difference (F (1,
109) D 11.06, p < .01, �2p D .10), which suggests that participants perceived a news story that was inconsistent with their opinions to be less biased when it was attributed to a congruent source (M D 4.33, SD D .17) than when it was attributed to an incongruent source (M D 5.27, SD D .19). Therefore, Hypothesis 2a was supported.

Hypothesis 2b proposed that audiences would perceive a congruent
news story as less biased when it came from a source that was consistent with
their own position than when it came from a source that was inconsistent
with their position. The results of a one-way ANOVA confirmed a significant
main effect of whether a participant’s opinion is consistent with the source’s

216 M. Kim

position (F (1,116) D 5.35, p < .05, �2p D .04). Participants considered a congruent news story to be less biased when it was delivered by a congruent source (M D 3.36, SD D .21) than when it came from an incongruent source (M D 4.03, SD D .21). Thus, Hypothesis 2b was supported.

Hypothesis 3 proposed that value-relevant involvement would have a
greater effect on the perceived bias of clearly slanted news than outcome-
relevant involvement. A hierarchical regression was conducted in order to
test Hypothesis 3. Demographics, content (congruent vs. incongruent), and
source (congruent vs. incongruent) were entered in the first block, and
value-relevant involvement and outcome-relevant involvement were entered
in the second block (see Table 1). In this regression model, content and
source (Block 1) were the manipulated factors, and the two types of in-
volvement (Block 2) were the measured factors. Collinearity statistics (value-
relevant involvement: tolerance D .92, Variance Inflation Factor (VIF) D
1.09; outcome-relevant involvement: tolerance D .97, VIF D 1.03) indicated
that the two involvement variables could be entered into the regression
without any collinearity concerns. The results of the regression showed
that neither value-relevant involvement (ˇ D �.03, p D .71) nor outcome-
relevant involvement (ˇ D �.02, p D .81) was related to the participants’
bias perceptions. Therefore, Hypothesis 3 was not supported.

Research Question 1 explored whether an audience’s value-relevant
involvement in a news topic would moderate the source effects on the bias
perceptions of the slanted news. A moderation test was conducted by using
the Hayes PROCESS macro (Hayes, 2013), treating sources (congruent vs.
incongruent) as an independent variable and value-relevant involvement as
a moderator. The interaction between source (incongruent was coded as
high) and value-relevant involvement was significant (b D .43, SE D .16, t D
2.67, p < .01). In order to probe this interaction, values of the moderator were sorted into three groups using a sample mean, as well as plus and minus

TABLE 1 Hierarchical Regression with Bias Perception as a
Dependent Variable

Predictor Model 1 Model 2

Gender �.067 �.071
Age �.055 �.048
Source .239* .238*
Content .328* .326*
Value-relevant involvement �.026
Outcome-relevant involvement �.016

R2 .179* .179*
Adjusted R2 .164 .157

Note. Standardized coefficients are reported.

*p < .001.

Partisan Source and Involvement 217

one standard deviation from the mean. The three groups were represented
as low, moderate, and high on the moderator. Figure 1 shows that the
interaction between source and value-relevant involvement approaches a
transverse interaction because the regression lines for the congruent and
incongruent sources move in opposite directions (Eveland, 1997). An effect
of source on participants’ perceived bias was not significant at low levels of
value-relevant involvement (�1 SD from the mean) (b D .23, SE D .27, t D
.84, p D .40), but the effect was significant at moderate (mean) (b D .81,
SE D .21, t D 3.84, p < .01) and high levels of value-relevant involvement (C1 SD from the mean; b D 1.39, SE D .33, t D 4.24, p < .001). The results showed that an effect of source (congruent vs. incongruent) on perceived bias of a news article was only manifest among those with moderate or high levels of value-relevant involvement in the focal issue of the article.

Research Question 2 explored whether an audience’s outcome-relevant
involvement in a news topic would moderate the source effects on the bias
perceptions of the slanted news. A second moderation test was conducted
by treating source as an independent variable and outcome-relevant involve-
ment as a moderator. The results indicated that there was no significant
interaction between source (incongruent was coded high) and outcome-
relevant involvement (b D �.27, SE D .15, t D �1.77, p D .08). The results
showed that outcome-relevant involvement did not moderate the source
effects on participants’ bias perceptions of the slanted news.

FIGURE 1 Conditional effects of congruent versus incongruent sources on bias perception
among those who have relatively low, moderate, or high levels of value-relevant involvement.

218 M. Kim

DISCUSSION

The present study explored the role of partisan sources (congruent vs. in-
congruent) in audiences’ bias perceptions of slanted news (congruent vs.
incongruent). Particularly, source effects were examined in the context of the
audiences’ two types of involvement, value-relevant and outcome-relevant
involvement.

Consistent with findings of prior research on the relative HME, the
interaction between news valence and the audiences’ opinions had great
effects on the audiences’ bias perception of slanted news stories. Audiences
with opinions that were incongruent with the news valence perceived news
stories from partisan sources as more biased than those with opinions that
were congruent with the news valence did. Source effects on perceived bias
of slanted news stories were also found. Audiences rated the news article,
regardless of its valence (congruent vs. incongruent), as less biased when
it was attributed to a source that was consistent with their own position
than when it came from a source that was inconsistent with their position.
Contrary to the expectation, all types of audiences’ involvement in the focal
issue were not related to the audiences’ bias perceptions of the slanted news
stories. While value-relevant involvement moderated the effects of partisan
sources on the bias perceptions of the slanted news stories, outcome-relevant
involvement did not. Source effects were only significant among those with
moderate or high levels of value-relevant involvement in the issue.

The results of this study have several practical and theoretical impli-
cations. First, this study suggests that partisan sources might amplify or
attenuate the relative HME. According to the relative HME, audiences with
prior opinions on an issue consider news stories that are incongruent with
their own opinions to be biased (Gunther & Chia, 2001; Gunther & Christen,
2002; Gunther et al., 2001). This bias perception could inhibit audiences
from ‘‘giving reasonable consideration’’ (Gunther et al., 2001, p. 316) to
news coverage that is inconsistent with their points of view, which enhances
opinion polarization regarding controversial issues. The results of this study
indicate that audiences perceive opinion-incongruent content that comes
from a source that is consistent with their own position to be relatively
unbiased. This suggests that if partisan sources continue to provide one-sided
content that only reflects their political position (conservative or liberal),
this may amplify the relative HME. In contrast, if partisan sources deliver
content whose valence is inconsistent with the sources’ existing political or
ideological disposition (e.g., when a conservative news outlet reports a news
story that contains a few liberal perspectives), this may mitigate the relative
HME.

The role of partisan sources in reporting controversial issues is particu-
larly important in South Korea. South Korea has a different journalism history
than that of Western societies (Oh & Park, 2005). For example, the institution

Partisan Source and Involvement 219

of journalism in the United States adopted a norm of objective reporting in
the turn of the twentieth century (Schudson, 2001). Although partisan media
have emerged particularly on cable news and the Internet, most mainstream
news outlets in the United States are considered to be objective and balanced
(Levendusky, 2013). In contrast, mainstream news outlets in South Korea are
developed based on their political or ideological positions, rather than on
professional ethics such as objectivity (Oh & Park, 2005). As a result, South
Korea’s mainstream news outlets are sharply polarized and only focus on
their political positions (conservative vs. liberal).

Ideologically polarized mainstream news organizations have been criti-
cized for aiding opinion polarizations in South Korea. In this situation, it is
necessary for news outlets to address this criticism by sharing a professional
norm of objectivity and fairness. This study suggests that news sources
that provide multiple perspectives, rather than only providing their parti-
san views, offer audiences opportunities to give consideration to alternate
perspectives of a controversial issue, which may contribute to the attenuation
of opinion polarization within that issue.

Second, this study extends the discussion of the effects of source on the
HME by examining the moderating functions of audiences’ different types
of involvement. Two types of involvement, value-relevant and outcome-
relevant involvement, were often intertwined in prior studies, leading to
theoretical ambiguity in understanding the relationship between audiences’
involvement and the source effects. This study attempts to reduce this theo-
retical ambiguity by exploring the distinct roles of different types of involve-
ment in the effects of source on the relative HME.

Unlike previous studies, which reveal that value-relevant involvement
‘‘leads people to process media coverage as more biased’’ (Choi et al., 2009,
p. 70), this study did not find any direct impacts of value-relevant involve-
ment on the perceived bias of slanted news. Value-relevant involvement
only influenced the bias perceptions by moderating the source effects. Those
with moderate or high levels of value-relevant involvement were significantly
influenced by partisan sources (congruent vs. incongruent), whereas those
with low levels of value-relevant involvement were not vulnerable to such
source effects. This is inconsistent with what social judgment theory predicts.
Social judgment theory suggests that because people with high levels of
value-relevant involvement have a wider latitude of rejection, they perceive
information to be more biased or hostile. Moreover, the hostile perceptions
of those with high levels of value-relevant involvement are not expected
to be easily influenced by source characteristics (Sherif et al., 1965; Sherif
& Sherif, 1967). Several attitude change studies (e.g., Johnson & Scileppi,
1969; Rhine & Severance, 1970; Sereno, 1968) also suggest that for those
with high value-relevant involvement, the sources of messages may have
little effect on attitude changes. In contrast, this study found that those with
high levels of value-relevant involvement evaluated news coverage to be

220 M. Kim

less biased or hostile when the coverage was attributed to a source that
was congruent with their own opinion. These findings provide confirmatory
evidence for the self-categorization explanation of the HME. A recent study
(Reid, 2012) suggests that self-categorization theory, rather than social judg-
ment theory, can explain the effects of source on the HME. According to
self-categorization theory, audiences’ internalized group membership (e.g.,
Democrats vs. Republicans) is a basis for their bias perceptions of the news
(Turner, Hogg, Oakes, Reicher, & Wetherell, 1987). Therefore, a news story
from an ingroup source is perceived to be less biased than the same news
from an outgroup source. When the position of individuals within a group
is more extreme, it is more likely that they will use group membership as
a critical element of their bias perceptions of the news (Reid, 2012). The
current study extends prior research on the self-categorization explanation
of the HME by focusing on the involvement types that underlie audiences’
group membership. As noted earlier, group membership can represent either
value-relevant or outcome-relevant involvement. The results of this study
show that only value-relevant involvement that underlies group member-
ship moderates the effects of source on the HME. Moreover, this study
demonstrates that the self-categorization explanation of the HME makes
sense under certain conditions of audiences’ value-relevant involvement.
Only those with moderate or high levels of value-relevant involvement seem
to categorize congruent and incongruent sources as ingroup and outgroup
sources, respectively. This means that self-categorization theory may only
explain the source effects (ingroup vs. outgroup) for people with high levels
of value-relevant involvement in an issue.

Third, because those who show greater value-relevant involvement tend
to hold a more fixed and polarized position on an issue, they are less likely
to be persuaded than those who show less involvement ( Johnson & Eagly,
1989). Contrary to this information, the current study suggests that if an
outgroup message is delivered by an ingroup source, it might be highly
persuasive for those with high-levels of value-relevant involvement. For
example, if ingroup news sources (conservative or liberal) present messages
from all of the sides fairly, including counterattitudinal messages, when
reporting on controversial subjects, this would provide their audiences with
a chance to rethink and understand the other party. An outgroup message
from an ingroup news source may have an especially striking effect on
changing attitudes of people with high ego involvement, who play a leading
role in opinion polarization within an issue by joining advocacy groups and
collective actions (Grunig, 1989).

The present study is subject to several limitations. This study did not
recruit participants from existing group memberships, instead grouping par-
ticipants based on their prior opinions (pro vs. con) on an issue. If future
studies examine source effects in the context of actual group memberships,
as many HME studies have done, this would be more helpful in under-

Partisan Source and Involvement 221

standing the role of participants’ social identity in their bias perceptions. The
issue that was used in this study, the dissolution of the Unified Progressive
Party, is more likely to be related to participants’ values rather than their
everyday lives. Therefore, it is required to verify the findings of this study by
using highly outcome-relevant issues. This study excluded participants with
a neutral position on the issue: In this study, 46.79% of participants who were
initially recruited had indicated that they held a neutral position on the issue.
Given the large portion of the ambivalent public, it is necessary to examine
how those who have a neutral position perceive slanted news and how
partisan sources influence their bias perceptions. In this study, participants’
bias perceptions were measured as either neutral or hostile. The direction
of the perceived bias was not precisely measured. If future studies explore
audiences who have a neutral position on an issue, a directional bias measure
would be more beneficial.

In conclusion, this study found different effects of source (congruent
vs. incongruent) on the HME and identified audiences’ value-relevant in-
volvement as a moderator of such effects. This study expands self-categori-
zation explanations for the HME, providing strong evidence that the self-
categorization of sources (ingroup vs. outgroup) plays a role in the bias
perceptions of news stories about a controversial issue only for audiences
with moderate or high levels of value-relevant involvement in that issue. This
study also provides empirical evidence that outgroup message from ingroup
source may attenuate the HME. These findings highlight the potential role
of partisan news outlets in a divided society where partisan audiences are
kept apart. Partisan news outlets may contribute to mitigate partisanship
or opinion polarization in the society by providing multiple perspectives
beyond their own ideological position.

REFERENCES

Ariyanto, A., Hornsey, M. J., & Gallois, C. (2007). Group allegiances and perceptions
of media bias: Taking into account both the perceiver and the source. Group
Processes & Intergroup Relations, 10, 266–279. doi:10.1177/1368430207074733

Arpan, L. M., & Raney, A. A. (2003). An experimental investigation of news source
and the hostile media effect. Journalism & Mass Communication Quarterly, 80,
265–281. doi:10.1177/107769900308000203

Chaiken, S. (1980). Heuristic versus systematic information processing and the use
of source versus message cues in persuasion. Journal of Personality and Social
Psychology, 39, 752–766. doi:10.1037/0022-3514.39.5.752

Cho, H., & Boster, F. J. (2005). Development and validation of value-, outcome-, and
impression-relevant involvement scales. Communication Research, 32, 235–
264. doi:10.1177/0093650204273764

Choi, J., Park, H. S., & Chang, J. J. C. (2011). Hostile media perception, involvement
types, and advocacy behaviors. Journalism & Mass Communication Quarterly,
88, 23–39. doi:10.1177/107769901108800102

222 M. Kim

Choi, J., Yang, M., & Chang, J. (2009). Elaboration of the hostile media phenomenon:
The roles of involvement, media skepticism, congruency of perceived media
influence, and perceived opinion climate. Communication Research, 31, 54–
75. doi:10.1177/0093650208326462

Christen, C. T., Kannaovakun, P., & Gunther, A. C. (2002). Hostile media perceptions:
Partisan assessments of press and public during the 1997 United Parcel Service
strike. Political Communication, 19, 423–436. doi:10.1080/10584600290109988

Eveland, W. P. (1997). Interactions and nonlinearity in mass communication: Con-
necting theory and methodology. Journalism & Mass Communication Quar-
terly, 74, 400–416. doi:10.1177/107769909707400211

Giner-Sorolla, R., & Chaiken, S. (1994). The causes of hostile media judgments. Jour-
nal of Experimental Social Psychology, 30, 165–180. doi:10.1006/jesp.1994.1008

Grunig, J. E. (1989). Sierra Club study shows who become activists. Public Relations
Review, 15, 3–24. doi:10.1016/S0363-8111(89)80001-3

Gunther, A. C., & Chia, S. C. (2001). Predicting pluralistic ignorance: The hostile
media perception and its consequences. Journalism & Mass Communication
Quarterly, 78, 688–701. doi:10.1177/107769900107800405

Gunther, A. C., & Christen, C. T. (2002). Projection or persuasive press? Contrary
effects of personal opinion and perceived news coverage on estimates of pub-
lic opinion. Journal of Communication, 52, 177–195. doi:10.1111/j.1460-2466.
2002.tb02538.x

Gunther, A. C., Christen, C. T., Liebhart, J. L., & Chia, S. C. (2001). Congenial public,
contrary press, and biased estimates of the climate of opinion. Public Opinion
Quarterly, 65, 295–320. doi:10.1086/322846

Gunther, A. C., & Liebhart, J. L. (2006). Broad reach or biased source? Decomposing
the hostile media effect. Journal of Communication, 56, 449–466. doi:10.1111/
j.1460-2466.2006.00295.x

Gunther, A. C., Miller, N., & Liebhart, J. L. (2009). Assimilation and contrast in a test
of the hostile media effect. Communication Research, 36, 747–764. doi:10.1177/
0093650209346804

Gunther, A. C., & Schmitt, K. (2004). Mapping boundaries of the hostile media effect.
Journal of Communication, 54, 55–70. doi:10.1111/j.1460-2466.2004.tb02613.x

Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process
analysis: A regression-based approach. New York, NY: Guilford.

Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communi-
cation effectiveness. Public Opinion Quarterly, 15, 635–650. doi:10.1086/266350

Iyengar, S., & Hahn, K. S. (2009). Red media, blue media: Evidence of ideological se-
lectivity in media use. Journal of Communication, 59, 19–39. doi:10.1111/j.1460-
2466.2008.01402.x

Jamieson, K. H., Hardy, B., & Romer, D. (2007). The effectiveness of the press in
serving the needs of American democracy. In K. H. Jamieson (Ed.), A republic
divided: The Annenberg Democracy Project (Pp. 21–51). New York, NY: Oxford
University Press.

Johnson, B. T., & Eagly, A. H. (1989). Effects of involvement on persuasion: A meta-
analysis. Psychological Bulletin, 106, 290–314. doi:10.1037/0033-2909.106.2.290

Johnson, B. T., Lin, H. Y., Symons, C. S., Campbell, L. A., & Ekstein, G. (1995). Initial
beliefs and attitudinal latitudes as factors in persuasion. Personality and Social
Psychology Bulletin, 21, 502–511. doi:10.1177/0146167295215008

Partisan Source and Involvement 223

Johnson, H. H., & Scileppi, J. A. (1969). Effects of ego-involvement conditions on at-
titude change to high and low credibility communicators. Journal of Personality
and Social Psychology, 13, 31–36. doi:10.1037/h0027992

Kim, K. S. (2010). Public understanding of the politics of global warming in the news
media: The hostile media approach. Public Understanding of Science, 20, 1–
17. doi:10.1177/0963662510372313

Levendusky, M. (2013). Partisan media exposure and attitudes toward the opposition.
Political Communication, 30, 565–581. doi:10.1080/10584609.2012.737435

Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude
polarization: The effects of prior theories on subsequently considered evidence.
Journal of Personality and Social Psychology, 37, 2098–2109. doi:10.1037/0022-
3514.37.11.2098

Oh, T. S., & Park, S. H. (2005). Hostile media perception: Message or messenger?
Korean Journal of Journalism and Communication Studies, 49, 135–290.

Reid, S. A. (2012). A self-categorization explanation for the hostile media effect.
Journal of Communication, 62, 381–399. doi:10.1111/j.1460-2466.2012.01647.x

Petty, R. E., & Cacioppo, J. T. (1979). Issue involvement can increase or decrease per-
suasion by enhancing message-relevant cognitive responses. Journal of Person-
ality and Social Psychology, 37, 1915–1926. doi:10.1037/0022-3514.37.10.1915

Petty, R. E., Cacioppo, J. T., & Goldman, R. (1981). Personal involvement as a
determinant of argument based persuasion. Journal of Personality and Social
Psychology, 41, 847–855. doi:10.1037/0022-3514.41.5.847

Rhine, R. J., & Severance, L. J. (1970). Ego-involvement, discrepancy, source cred-
ibility, and attitude change. Journal of Personality and Social Psychology, 16,
175–190. doi:10.1037/h0029832

Sereno, K. K. (1968). Ego-involvement, high source credibility, and response to a
belief-discrepant communication. Speech Monographs, 35, 476–481.

Sherif, C. W., & Sherif, M. (1967). Attitude, ego-involvement, and change. New York,
NY: Wiley.

Sherif, C. W., Sherif, M., & Nebergall, R. E. (1965). Attitude and attitude change:
The social judgment-involvement approach. Philadelphia, PA: Saunders.

Schudson, M. (2001). The objectivity norm in American journalism. Journalism, 2,
149–170.

Sundar, S. S., & Nass, C. (2001). Conceptualizing sources in online news. Journal of
Communication, 51, 52–72. doi:10.1111/j.1460-2466.2001.tb02872.x

Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987).
Rediscovering the social group: A self-categorization theory. Oxford, UK: Black-
well.

Vallone, R. P., Ross, L., & Lepper, M. R. (1985). The hostile media phenomenon:
Biased perception and perceptions of media bias in coverage of the Beirut mas-
sacre. Journal of Personality & Social Psychology, 49, 577–587. doi:10.1037/0022-
3514.49.3.577