Political Behavior

, Volume 29, Issue 4, pp 441–464

The Messenger Overwhelming the Message: Ideological Cues and Perceptions of Bias in Television News


    • Department of Political ScienceWestern Kentucky University
Original Paper

DOI: 10.1007/s11109-007-9031-z

Cite this article as:
Turner, J. Polit Behav (2007) 29: 441. doi:10.1007/s11109-007-9031-z


Survey research has demonstrated that citizens perceive ideological bias in television news, specifically with regard to CNN and Fox News Channel (FNC), which allegedly represent the liberal and conservative viewpoint, respectively. In this paper I argue that attaching the CNN and FNC labels to news stories sends an ideological cue to the viewer regarding the content of the story. Utilizing an experimental design that allows manipulation of the network attribution of actual FNC and CNN content, I am able to demonstrate that the CNN and FNC labels function as ideological signals to the viewer, with this signal being most pronounced among ideologues whose views are supposedly at odds with those attributed to the network.


Television newsIdeological biasHeuristics


Public opinion surveys have consistently demonstrated that a large number of citizens perceive an ideological bias within television news.1 For many Americans the two television news outlets that epitomize this bias are CNN and Fox News Channel (FNC), which according to Nielsen ratings are the two most popular 24-hour news networks.2 While surprisingly little work has been done on this issue, preliminary evidence demonstrates that at least 35% of the American public perceives FNC as being overtly conservative and at least 26% perceive CNN as being liberally biased.3

Given the well-established role of the news media as political informant to the American public, it is important to evaluate whether this perceived ideological bias influences how people process television news. Clearly, there is a real risk that a widespread view among citizens that these media outlets are ideologically biased could pose a substantial obstacle to informing the American public about politics. It may be the case that attaching the “CNN” and “FNC” labels to news stories is virtually the equivalent of sending ideological signals to the viewer. Knowing a story is from one of those networks may send signals to the viewer regarding the ideological nature of the news content, which could lead the viewer to perceive ideological bias as well as raise cognitive roadblocks that prevent him or her from properly acquiring and processing the information. Therefore, the goal of this study is to gain a deeper understanding of information processing as it relates to information received from the televised news media.

Heuristic Processing and Television News Interpretation

There is certainly disagreement on whether ideological bias exists in the news media and what form it takes. A survey conducted by the Program on International Policy Attitudes (PIPA) (2003) demonstrated that FNC viewers were more likely to hold misperceptions about the Iraq War and the War on Terror, which has resulted in allegations that FNC pushes a conservative agenda. In contrast, Groseclose and Milyo (2005) assert that nearly all of the news media are liberally biased. Using a technique wherein they calculate scores on the ADA metric for news media outlets, the authors argue that the news media, with the exception of FNC’s Special Report with Brit Hume and The Washington Post, push a liberal agenda.

Although these studies are indeed provocative, they have been subjects of scrutiny. The authors of the PIPA study are unable to demonstrate that misperceptions about the Iraq War and the War on Terror emerged because falsehoods were reported by FNC. Therefore, as is noted in the PIPA report, it is premature to assume causation (and thus ideological bias) based on this study when only a correlation has been demonstrated. It is entirely plausible that characteristics of viewers, not of the news being viewed, drove the observed variance in misperceptions. In addition, the Groseclose and Milyo study has received criticism due to potential problems with methodology and coding choices.4

Despite the findings of a few recent studies, the vast majority of previous scholarly research provides reason to believe that the actual content of television news broadcasts is not ideologically biased (Bennett & Entman, 2000; Entman, 1989; Graber, 1980; Iyengar, 1991; Iyengar & Kinder, 1987; Robinson & Clancey, 1985; Robinson & Sheehan, 1983; Weaver, 1972,). However, irrespective of what most of this previous research has demonstrated, a substantial portion of the American public still perceives CNN and FNC as being ideologically biased. Hence, we are left with an intriguing discrepancy. Systematic academic research struggles to find more than a hint of ideological bias in the news, yet the perception that such bias exists is widespread among the mass public. The critical assumption motivating the current research is that this perception of ideological bias can be highly consequential irrespective of whether ideological bias exists in reality. For the viewer who perceives CNN and FNC as advancing liberal and conservative views, respectively, those networks’ labels function as powerful ideological cues. Therefore, simple attribution of news content to CNN or FNC can activate a cognitive heuristic, shaping how given stories are perceived. Importantly, provided that viewers assume ideological bias to be present, this heuristic process can be activated irrespective of whether the actual content of the news is ideologically biased. In short, perception may trump reality.

Research on cognitive heuristics has been developed largely with reliance upon dual-process models of information processing. A notable theoretical model in this genre is Petty and Cacioppo’s (1986) Elaboration Likelihood Model, wherein the authors argued that individuals process information via either a central or a peripheral route. Central route processing refers to methods of evaluation in which the individual gives careful consideration to the information presented when making a decision. By contrast, individuals making decisions via the peripheral route make judgments based on simplistic reactions to stimuli that are, in many cases, only tangentially related to the message at best. One of these peripheral methods involves the use of heuristic processing (Chaiken, 1980, 1987; Chaiken, Liberman, & Eagly, 1989; see also Tversky & Kahneman, 1974). According to Sniderman, Richard, and Philip (1991), heuristics are efficient tools for the decision maker to utilize because they require little information while providing answers to complex problems. Essentially, reliance on heuristics allows the individual to evaluate information without having to engage the information in an in-depth manner. Because most individuals deal with matters that are of much greater importance to them than analyzing politics, it is highly likely that individuals employ heuristics when trying to understand political news.

In this case viewers appear to rely on two complementary heuristics when interpreting a signal that they attribute to either CNN or FNC. The first of these, the ideological heuristic, speaks to the content of the news. Early social psychology research demonstrated that individual motives, such as ideology, can shape the way individuals perceive the things they see (Hastorf & Cantril 1954). In the field of political science, Joslyn and Ceccoli (1996) demonstrated that the effectiveness of media messages relies on the relationship between viewer ideology and the valence of network news coverage. Entman (1989) and Dalton, Beck, and Huckfeldt (1998) found that perception of news is shaped as much by a person’s political views as by objective content. Scholarly work on hostile media bias, wherein those with entrenched political beliefs tend to believe that the media are hostile toward their political position, also reinforces this idea (Vallone, Ross, & Lepper, 1985). As a result, ideology appears to be a strong, consistent influence on the interpretation of news broadcasts.

The second heuristic at work here is source cues. Source cues should be significant because they provide an easy way for an individual to evaluate a message (Carmines & Kuklinski, 1990; Mondak 1993a, b). In an early study, McGuire (1969) concluded that people evaluate a candidate and either accept or reject that candidate’s message based solely on a heuristic evaluation. In addition, Zaller (1992) demonstrated that when people process information they use the reputation of the message provider as a cue for evaluating the message. The work of Riggle, Ottati, Wyer, Kuklinski, and Schwarz, (1992), Ottati (1990), and Ottati, Fishbein, and Middlestadt (1988), Ottati and Wyer (1990) has also supported these findings.

Kuklinski and Hurley (1994) connected the use of ideological heuristics and source cues and, as a result, highlighted one area of possible peril when citizens draw on simple signals to formulate political evaluations. They argued that by focusing their attention on the individual political actor, citizens make quick judgments of the information presented to them based largely on the reputation of the speaker. An easy way to characterize this is, because respondents are evaluating information based on their opinion about the speaker rather than an evaluation of the content presented, the "messenger" is overwhelming the "message". They found that this was the case, as experimental subjects presented with a message evaluated that message based largely on their opinion of the speaker. In this case, since both networks arguably send strong cues to the viewer regarding the ideological slant of their coverage, the actual “message” may be irrelevant so long as the viewer knows that the “messenger” was either CNN or FNC.

I believe that the “messenger” may indeed be overwhelming the “message” due to viewers’ ideological perception of CNN and FNC. By assessing whether this is the case, we can hopefully draw firm conclusions as to whether citizens are indeed drawing ideological cues from these networks, which, irrespective of story content, leads to an overall perception of ideological bias by viewers.


The two hypotheses for this examination are as follows:

Hypothesis 1

Irrespective of the substantive content of news stories, attribution of those stories to FNC and CNN will cause viewers to perceive the stories as ideologically biased.

I expect a significant difference in viewers’ perception of ideological bias between the networks, with news content attributed to FNC being perceived as conservatively biased and news content attributed to CNN being perceived as liberally biased.

Hypothesis 2

Perceptions of bias will be most pronounced among ideologues whose views are at odds with those purportedly held by a given network.

The tendency to view news content attributed to FNC as conservatively biased should be most pronounced among liberals. Likewise, the tendency to view news content attributed to CNN as liberally biased should be most pronounced among conservatives.

Experimental Design

I anticipate that people evaluate ideological bias in newscasts as a reaction to the network cue as opposed to any actual ideologically biased content within the newscasts. To test my hypotheses I employed a multistage experimental research design to determine whether the messenger (network) is overwhelming the message (substantive content of the news).5 An experimental design is employed because it allows the researcher to create the necessary experimental condition rather than hoping that the condition would eventually occur in the real world (Iyengar, Peters, & Kinder, 1982). In this instance an experimental design is extremely beneficial because, in the real world, there is a perfect correlation between the airing of a substantive news story by CNN (or FNC) and the presence of the CNN (or FNC) cue. Investigating whether these network attributions are indeed sending ideological cues to the viewer would be impossible because there would be no means to determine if viewers’ perceptions were driven by the substantive content of the news, the ideological signals transmitted via the network cue, or a combination of the two. However, an experimental setting allows me to separate network attribution from content and investigate my theoretical expectations.

In this experiment I selected five stories from FNC and CNN, and then created replications of the stories wherein the content remained constant but the network attribution was manipulated. The core of the design is a two by three factorial, which resulted in the creation of six treatment groups. There are six treatment groups because the actual news content came from either CNN or FNC, and the content was attributed to either CNN, FNC, or given no network attribution. The first stage of this process involved deciding which stories from FNC and CNN to present to my research subjects. I videotaped several hours of news footage from CNN and FNC over a three day period in January 2004. Then, I reviewed the videotaped footage and selected five stories from each network for this analysis. It was first necessary to eliminate stories that would not work within the confines of this experiment. Two mechanisms were used to achieve this end. First, stories that featured highly recognizable “talking head pundits” such as FNC’s Sean Hannity or Bill O’Reilly and CNN’s James Carville or Paul Begala were not used. These pundits present “news analysis”, whereas my concern here is whether the actual hard news stories aired by CNN and FNC convey ideological cues. Second, stories that contained a substantial amount of back and forth banter between the reporter and anchor were not chosen because of the logistical difficulties of replicating the anchor/reporter interaction.

The final selection of stories came from those left over after the initial elimination process. This was admittedly a subjective decision, as there were stories that were not chosen that could have easily been substituted for the ones that were chosen. Those ultimately chosen were ones that I believe closely mimicked the coverage one would see on a broadcast during the period in which this study was conducted (stories that cover domestic politics, foreign affairs, health concerns, human interest, etc.).6 The five CNN stories that were chosen were about the war in Iraq, nuclear inspections in North Korea, a female suicide bomber in Israel, a newly proposed “support of marriage” initiative, and Saddam Hussein. The five FNC stories that were chosen were about the war in Iraq, nuclear inspections in Libya, mad cow disease, illegal aliens, and Saddam Hussein.

The second stage of this experimental design required the creation of believable replications of these original news broadcasts. Creating believable replications was essential in order to maximize the validity of this study. To achieve this end, the replications were created at a professional television studio, relying on the assistance of a professional news anchor with 23 years of television news broadcast experience. In addition, using a professional television studio gave the videos the appearance of a real news broadcast and thus lessened the artificiality that commonly plagues experimental research. The anchor read transcripts from the original FNC and CNN stories in order to replicate the exact language used in the stories. The experimental manipulations were achieved by the broadcaster vocally identifying with which news organization he was affiliated.7 The broadcaster recorded each set of stories from CNN three times: once identifying himself as being affiliated with CNN, once identifying himself as being affiliated with FNC, and once giving no network affiliation. Likewise, the broadcaster recorded the set of stories from FNC three times: once identifying himself as being affiliated with CNN, once identifying himself as being affiliated with FNC, and once giving no network affiliation. Each replication set had a running time of roughly ten minutes.

It is important to note that any artificiality that remains in the replications should constitute a conservative bias for hypothesis testing. Although the news stories were produced in a professional news studio, the studio lacked the characteristic trappings of the major cable news networks. Also, network affiliations were conveyed via verbal reference to the networks, but these references were not reinforced with the network news logos. In short, FNC and CNN broadcasts make it even clearer what network the viewer is watching, and the fact that my replications could only approximate a portion of this cueing effect means that any ideological signals transmitted in my tests almost certainly were weaker than those conveyed by actual FNC and CNN broadcasts.

The next stage in this study consisted of actually conducting the experiment and collecting the data. A sample of 246 undergraduates enrolled in political science courses at a major state university participated in this experiment in the spring of 2004. Participants for the study were recruited from various introductory and upper-level undergraduate political science courses. When recruited participants were simply told that they would be participating in a media focus group. Participants were offered extra credit in the course they were enrolled in as an incentive to take part in the study.8 Those who volunteered initially had the ability to opt out of the study (i.e., to not show up for their assigned viewing time). However, those respondents who, for whatever reason, failed to appear, did not receive extra credit.

Participants were randomly assigned to one of the aforementioned experimental conditions. This was achieved by matching the name of a randomly selected volunteer to a randomly chosen experimental treatment. In this instance, randomization was successful, as the experimental conditions were roughly equal with regard to the gender, party identification, and political ideology of respondents. This randomization process also ensured that the size of each experimental cell was roughly the same.

After viewing the news stories, participants completed a survey instrument. This instrument contained questions that achieved a number of goals. The most important aspect of the survey was that it provided a measure of the dependent variable under examination. It also provided a substantial amount of potentially useful background information by administering a battery of questions regarding socio-demographic and personality characteristics. The survey instrument was not administered until after the subjects had viewed the entire video, both so respondents could devote their full attention to the videotapes and to avoid signaling to the respondents what they should specifically be looking for in the videotapes in order to answer the questions.

The survey served two valuable purposes in addition to providing measures of the study’s variables of interest. First, the survey helped mask the actual purpose of the study. If you recall, participants were told that they were participating in a media focus group. By including questions related to that activity I was able to hide the actual purpose of the study within a broad survey instrument. Only after conducting the experiments were respondents alerted to the more specific goals of the study.

Also, as a check to make sure respondents were not keying in to what the study was actually about, I concluded the survey by asking them to make a guess as to the project’s specific purpose. There was no systematic answer to this question, as responses ranged from the cover story to the admission of no idea. Ten of the survey respondents correctly guessed some variant of “media bias study.” Because this response indicated the possibility that these respondents were “cued in” to the specific aims of the study, these ten respondents were removed from the analysis.

Variable Measures

The dependent variable for this analysis is the perceived ideological bias of the news provided by the viewer. This was measured in terms of how far away the viewer thought the story was from an ideological neutral point. In this study respondents viewed the five stories and then were asked to rank them on a 0–10 scale, with 0 representing strongly perceived liberal bias, 5 representing perceived ideological neutrality, and 10 representing strongly perceived conservative bias. Because I am interested in the overall evaluation of bias in the news and not just bias in individual news stories, the scores of each story were summed in order to get a summary measure of ideological bias within the news program.9 Therefore, bias is measured on a 0 to 50 scale, with 0 representing absolute perceived liberal bias, 50 representing absolute perceived conservative bias, and 25 representing perceived ideological neutrality.

There are several additional variables that need to be explained. The first of these is a measure of the overall perceived ideological bias of six major television news networks: CNN, FNC, ABC, CBS, MSNBC, and NBC. These variables were measured in the same manner as the ideological bias measure of the stories that were mentioned previously. Respondents were asked to rank the networks on a 0 to 10 scale, with 0 representing strongly perceived liberal bias, 5 representing perceived ideological neutrality, and 10 representing strongly perceived conservative bias.

Because hypothesis two deals with ideology, it was necessary to have a sound measure of that concept. Political ideology was initially measured on the traditional one (strong liberal) to seven (strong conservative) scale; for analytical reasons, this variable was recoded from negative three to three, with negative three signifying a “strong liberal”, zero signifying an “ideological moderate”, and three signifying a “strong conservative”.

In this analysis it was necessary to control for the attribution of the stories to determine if this alone was driving bias scores. Two dichotomous variables were created to account for whether the stories were attributed to CNN or FNC. The CNN attribution variable is coded zero if the stories are not attributed to CNN and one if they are. The FNC attribution variable is also coded zero if the stories are not attributed to FNC and one if they are. Because I am interested in how ideologues react to individual networks, it was necessary to create interaction terms. Political ideology was interacted with the FNC and CNN attribution dichotomous variables in order to determine how respondents harboring different ideologies respond to different network attributions. These two interaction terms allow for an assessment of differential ideological responses to news networks with allegedly different ideological viewpoints.

I also controlled for the original source of the story to make sure that nothing about the content from one network or the other was driving bias scores. This variable was coded zero if the original source was FNC and one if the original source was CNN. In addition, the original news story source variable was interacted with ideology.

Manipulation Checks

To draw valid conclusions from this study it was essential that the participants were responsive to the experimental manipulations. The first question is whether participants noted and accepted the experiment’s network attributions. After all, it would be meaningless to attribute news stories to CNN, FNC or any other network were it the case that viewers tune out when such network affiliations are mentioned.

To test this, I asked participants what network produced the stories they viewed. What we should find is that participants who viewed stories attributed to FNC and CNN accurately reported those attributions. Consistent with that expectation, 92.4% of the experimental subjects retained the experimental manipulation. Beyond serving as a manipulation check, this finding has two important substantive implications. First, the high rate of success for the manipulation indicates that it was not difficult to convince subjects that CNN content came from FNC and vice versa. Secondly, those who did not retain the manipulation were dropped from the analysis, because people who do not grasp or do not care what network they are watching should essentially be immune to any cueing effects from the network attribution.

Because one of the major assumptions underlying this study is that viewers perceive CNN and FNC as the most liberal and conservative news organizations, respectively, it is important that my sample views these two networks in that way.

Table 1 indicates that this is indeed the case. Keep in mind that respondents were asked to rank these networks after they were asked about their perceptions of the stories in order to ensure that their thoughts on the networks were not influencing their opinions regarding the stories. Also, in this analysis the respondents who received an attribution were separated from those who did not in order to control for the possibility that people may have been judging the networks based on the content of the stories they just viewed (i.e., respondent feels stories are too liberal, knows stories are from CNN, and therefore ranks CNN as being more liberal than he/she initially would have). Roughly one quarter of respondents viewed news stories without network attributions, and thus their perceptions of the ideological tone of the networks cannot be influenced by the experiment itself. It turns out that this distinction makes little statistical or substantive difference in the analysis. The reference point for these groups is a score of five, which represents perfect ideological neutrality for the network. In both portions of the analysis, participants perceived CNN as the most liberal and FNC as the most conservative of the major news organizations. These networks are also the only ones whose evaluations are significantly different from the point of ideological neutrality. This is substantively important because it justifies my choice of television news networks for this analysis.
Table 1

Perceived ideological bias of major television news networks


No attribution repsondents

All respondents






















Note: Superscripts indicate means that are significantly different from 5.00 (perfect ideological neutrality); * p < .05 ** p < .01 *** p < .001

Table 2

Perceptions of attributed stories, originals versus replications

Characteristic of broadcast



CNN N = 47

    Anchor appearance



    Anchor voice



    Anchor knowledge



    Anchor trust



    Anchor objectivity



    Story ideology



FNC N = 47

    Anchor appearance



    Anchor voice



    Anchor knowledge



    Anchor trust



    Anchor objectivity



    Story ideology



Note: * p < .05 ** p < .01

As a final manipulation check I compared the CNN and FNC originals with the replications to make sure there was not any discernible difference between the two that may be influencing any potential evaluation of ideological bias.

The original versions of these stories were compared with the replications on six different criteria: anchor appearance, anchor tone of voice, anchor level of knowledge, anchor trustworthiness, anchor objectivity, and story ideology.10 As Table 2 demonstrates, the only significant differences that emerge involve appearance and tone of voice. There is no significantly discernible difference in level of knowledge, trust, objectivity, and story ideology, between the originals and the replications. This is important for two reasons. First, this indicates that the replications were successful in capturing the ideological tone of the original broadcasts. Second, the fact that no difference exists in most categories (specifically story ideology) allows me to pool the originals with the replications that were attributed to the proper network and analyze them within the two by three factorial framework.


The model in Table 3 serves as the key test of hypothesis 1. I expect that news content attributed to CNN will have the smallest mean on the story ideology scale and that news content attributed to FNC will have the largest, and that there will be a significant difference between the two. A regression model was utilized to investigate if there is a difference in means across the three network attributions.
Table 3

Perceived ideological bias of television news by network attribution









CNN content




CNN attribution




FNC attribution




N = 218

F = 44.24

Prob > F = .0000

Adj. R2 = 0.374

p < .001

The findings presented in Table 3 provide support for my first hypothesis. Stories that are not attributed to either FNC or CNN register a score of 25.5 on my 0–50 story ideology scale, meaning that respondents essentially perceive these stories to be ideologically neutral. The negative coefficient for CNN attribution confirms my suspicion that viewers perceive stories reported on that network as having a liberal bias, as stories attributed to CNN are seen as being roughly 4.3 points more liberal than stories that receive no network attribution. Likewise, the positive coefficient for news stories attributed to FNC confirms that viewers perceive stories presented on that network as having a conservative tilt, as stories attributed to this network are seen as being roughly 4.9 points more conservative than stories that received no network attribution. The fact that both the FNC and CNN attribution variables are significant confirms that each of these attributions influences respondents’ rating of the stories presented. The insignificant content variable illustrates that there is no perception of an underlying ideological bias within the content of the stories presented by either network. Therefore, whether the original source is FNC or CNN, respondents are only responding to the network attribution when formulating opinions regarding ideological bias within news stories.

The model presented in Table 4 presents the key test for hypothesis 2.11 I expect conservative viewers to perceive news content attributed to CNN as being much more liberal than do moderates or liberals. Likewise, I anticipate that liberal viewers will perceive news content attributed to FNC as being much more conservative than do moderates or conservatives. Once again, a regression analysis was also utilized to investigate this hypothesis.
Table 4

Perception of ideological bias in television news by political ideology and network attribution









CNN content




CNN attribution




FNC attribution








Ideology X CNN attribution




Ideology X FNC Attribution




Ideology X CNN Content




N = 218

F = 36.514

Prob > F = .0000

Adj. R2 = 0.535

p < .05

Again, the results comport well with expectations. The following equations isolate the conditional nature of the relationship between network attribution and political ideology. The influence with regard to news content attributed to CNN is illustrated in Equation 1:

$$ {\text{EQ 1}}:{\text{ Influence of CNN Attribution}} = [ - 4.549 - 1.501{\text{ }}({\text{Ideology}})] $$

The significant coefficient of −4.549 operating on CNN Attribution is the conditional effect for ideological moderates, or when the value of ideology is set at 0. What this reveals is that merely attributing news content to CNN cues moderates to a liberal news bias, as this attribution moves this ideological group 4.549 points in a negative direction on the story ideology scale. The significant interaction indicates that ideology conditions the influence of attribution. If you will recall, the coding of liberals ranged from −3 to −1. Therefore, as respondents become increasingly liberal, or as the value of ideology decreases, the effect of attributing news content to CNN becomes less influential, ultimately reaching a point to where attribution to CNN has essentially no influence on strong liberals (i.e., strong liberals do not perceive ideological bias from content attributed to this network). On the other hand, coding of conservatives ranged from 1 to 3. As a result, as respondents become increasingly conservative, or as the value of ideology increases, the effect of attributing news content to CNN is exacerbated, as strong conservatives are more apt than any other ideological group to detect liberal bias in news content attributed to CNN.

The conditional relationship with regard to news content attributed to FNC is illustrated in Equation 2:

$$ {\text{EQ }}2:{\text{ Influence of FNC Attribution}} = [4.736 - 1.502{\text{ }}({\text{Ideology}})] $$

Once again, the significant coefficient of 4.736 operating on FNC attribution is the conditional effect for ideological moderates. This indicates that simply attributing news content to FNC cues moderates to a conservative news bias, as this attribution moves this ideological group 4.736 points in a positive direction on the ideological bias scale. In addition, the significant interaction indicates that ideology conditions the influence of FNC attribution. In this case, as respondents become increasingly liberal, the influence of attributing news content to FNC reaches its peak, as strong liberals are most likely of all ideological groups to detect ideological bias in news content attributed to this network. In contrast, for strong conservatives, attributing news content to FNC has little substantive effect.

Figure 1 demonstrates the substantive effect of these interaction terms. As the figure illustrates, for strong liberals, CNN attribution conveys nothing extraordinary, and thus strong liberals’ story ideology scores are nearly identical for content that received a CNN attribution and for the no attribution control. However, mere attribution of stories to FNC results in a stark nine-point jump on the dependent variable, with strong liberals seeing the FNC label as a signal of conservative bias. Precisely the opposite pattern emerges for strong conservatives. For this group, attributing stories to FNC brings no change relative to the no-attribution baseline, but story ideology rating decreases nine points when coverage is linked to CNN.
Fig. 1

Perceptions of ideological bias in television news by political ideology and network attribution

When we view the CNN and FNC effects in tandem, the impact of the ideological signals transmitted by these networks becomes most vivid. Strong conservatives give news attributed to CNN a story ideology rating of 13.89. Strong liberals give the same news attributed to FNC a story ideology rating of 35.98. Holding actual content constant, the mere attribution of that content to one network or the other causes an enormous disparity in perceived story ideology among these two ideological groups; a disparity which, at over 22 points, spans nearly half of the range of the story ideology scale.

In contrast to ideologues, who perceive networks with ideological reputations at odds with their own as biased, Figure 1 reveals that the CNN and FNC attributions both matter for moderates. Moderate viewers perceive news attributed to CNN and FNC as having a liberal and conservative bias, respectively. Therefore, when a network’s reputation is not at odds with a respondent’s ideological preferences, ideological bias is not perceived; otherwise, the mere linking of a story to FNC or CNN sends a powerful ideological signal to the viewer regarding the nature of the content.12

Conclusion and Implications

Ideological bias in the news is one of the more discussed yet least understood phenomena for observers of contemporary political science. My research has attempted to shed light on one important facet of this topic. I have argued that attaching the “CNN” and “FNC” labels to news stories sends an ideological signal to the viewer, which could lead the viewer to perceive ideological bias and to raise cognitive roadblocks that prevent the proper acquisition and processing of the information. My findings have largely supported this argument; at least in the case of CNN and FNC, the “messenger” does appear to be overwhelming the “message”.

Utilizing an experimental approach permitted me to hold news content constant, while manipulating network attribution. My results demonstrate that people do react to the label of the news organization to which the story is attributed. Irrespective of the actual content of the stories presented, the CNN and FNC labels send a clear signal to the viewer that the content of the reports is ideologically biased: to the liberal side for CNN and to the conservative side for FNC

Furthermore, these perceptions of bias are most pronounced among ideologues who believe that their views are most at odds with those of a network. Perceptions of liberal bias in CNN are most pronounced among strong conservatives, and perceptions of conservative bias in FNC are most pronounced among strong liberals. In addition, I was also able to demonstrate that these perceptions of bias were not lost among weaker ideologues, as even moderate viewers identified a liberal bias in news content attributed to CNN and a conservative bias in news content attributed to FNC.

What are the potential implications of the findings? First, it is clear that the perceptions of CNN and FNC as being ideologically biased are pervasive in society. This examination has clearly illustrated that individuals can watch virtually identical stories on CNN and FNC yet still derive different ideological signals from each network. These findings would appear to provide support for those on both sides of the aisle who allege that the media has either a conservative (Alterman, 2003, Gitlin, 2000; Miller, 2002) or liberal (Goldberg, 2001; Maitre, 1994) bias. However, the failure of respondents to identify ideological bias in instances where the news stories are not given a network attribution draws these allegations into question. Perhaps a more apt description of what occurs is that ideological bias in television news is merely an artificial construction that has become so ingrained in American culture that the organization presenting the news has become more important than the news product it presents. This is problematic because it could result in people dismissing unbiased news reports because they wrongly believe that they are ideologically biased. In the worst case, the misconception that CNN and FNC function as the news outlets of the liberal and conservative viewpoint, respectively, could ultimately lead to public confusion regarding what political ideals liberals and conservatives actually strive to uphold.

These findings also lend support to scholars who warn of the potential dangers of reliance upon heuristic shortcuts. At first glance one could argue that these findings support the notion, first put forward by Converse (1964), that the citizenry is woefully ignorant with regard to politics and, in this case, political news. However, one must remember that respondents were able to identify apparently neutral reports as ideologically neutral when not presented with a network attribution, which indicates that the citizenry does possess the capacity to engage information presented on a news broadcast. The problem emerges when citizens are also presented with the network attribution. This allows respondents to simplify and engage in peripheral processing, and in this case the reputation of these networks appears to lead citizens to have misperceptions about the ideological nature of the information presented. These misperceptions, or mistakes, in the evaluation of information that result when citizens rely on cognitive heuristics, is the “cautionary tale” that was told by Kuklinski and Hurley (1994) and supported by other scholars in both psychology (Tversky & Kahneman, 1974) and political science (Lau & Redlawsk, 1997; Mondak 1994).

The results presented here have interesting implications for other areas of research in the field of media and politics. Although nothing in my findings rejects the significance of phenomena such as priming, framing, and agenda setting, my results do indicate that those effects may vary widely across individuals depending on what type of ideological bias, if any, each person perceives in the news. As Miller and Krosnick (2000) pointed out, perceptions of bias translate into media distrust, and can possibly prevent the news media from successfully fulfilling its priming and agenda setting functions. Walsh (2003) makes a related point when she argues that media effects interact with individuals’ underlying political perspectives. If nothing else, future researchers in these areas can gain added nuance by recognizing that, for example, the capacity of CNN to produce an agenda setting effect may well depend in part on the ideological leaning of the particular viewer.

Although viewers can be assigned to particular viewing conditions in the laboratory, self-selection operates in the real world. One manifestation of self-selection is that some liberals avoid watching FNC, and some conservatives steer clear of CNN. These patterns warrant assessment in light of the current findings. It is certainly true that FNC has a more conservative viewership and that liberals are more likely to watch CNN, but it is not the case that liberals and conservatives fully abstain from FNC and CNN, respectively. Therefore, while self-selection may limit the effects on the margins, it certainly does not eliminate the effects. Furthermore, these individuals are likely self-selecting because they think they disagree with the network, as earlier I demonstrated that, absent network attribution, the news content presented in this experiment was perceived as being ideologically neutral. If anything, this adds another layer of support to my argument that network reputation, and only network reputation, matters with regard to perceptions of ideological bias in television news.

The next step in this research agenda is to examine what, if any, impact these perceptions of bias have on how the American citizen processes television news. Specifically, research needs to examine whether individuals are better able to acquire and retain information presented to them by a network they believe to be consonant with their ideological beliefs or by one that they believe to be ideologically dissonant. Compelling arguments can be made to support both theses. It seems plausible that someone would remember more information presented by a consonant media source, as it would serve to reinforce his or her existing beliefs. However, it is also plausible that people will remember more information presented to them by a dissonant source, because the sheer fact that they disagree with the source may serve as a motivating factor to cognitively engage the material.

In conclusion, does FNC have a conservative slant? Is CNN liberally biased? In one important sense, the answers do not really matter. So long as the viewers perceive ideological bias—and they do—allegations of ideological bias have the potential to produce very real consequences for American public opinion.


These surveys were conducted by the American Society of Newspaper Editors (1999), the Gallup Organization (2003), and the Pew Research Center (2005).


Although Fox News Channel has recently been on top in the competition for ratings, it is conceivable that either network could be in first place with the other in second place, depending on when one examines the ratings. The important point, however, is that both networks outpace their cable news competitors.


This poll was conducted on behalf of Media Matters for America by the Garin-Hart-Yang Research Group, Washington, D.C., in November 2004.


Alterman, Eric (2006). Think Again: Rigging the Numbers. Center for American Progress. http://www.americanprogress.org/issues/2006/01/b1347483.html


This research was conducted in accordance with the guidelines set forth by the Human Subjects Committee of the Office of Research at Florida State University. A detailed explanation of these guidelines can be found at www.research.fsu.edu/humansubjects.


The State of the News Media 2004 study, conducted by Journalism.org, confirmed that foreign policy stories were a major part of cable news broadcasts when this study was conducted, as this type of story constituted 24% and 21% of the broadcast time on CNN and Fox News Channel, respectively. However, this is not always the case, which raises questions regarding whether these results extend to all periods and forms of news coverage. Within the confines of my data, I can determine whether respondents reacted differently to domestic and foreign news stories. They did not, as the difference of means between foreign and domestic stories originating from CNN and Fox News was only .16 and .15, respectively, and not statistically significant. Therefore, it does not appear that the mix of domestic/foreign stories influences the nature of results one way or the other. However, additional attention is warranted regarding the possibility that perceptions of ideological bias differ for domestic vs. foreign news stories.


The anchor identified his network affiliation in three different ways. First, at the beginning of the broadcast, the anchor stated: “This is Phil Streetman reporting for CNN/Fox News.” Second, during the stories the anchor would use phrases that conveyed network affiliation without changing the tone of the story, such as “CNN/Fox News has learned...” or “sources have told CNN/Fox News...”. Finally, the anchor signed off using the network taglines: “This is Phil Streetman for CNN, the most trusted name in news” and “This is Phil Streetman for Fox News, where you always get fair and balanced coverage.”


Individual professors had control over the manner and amount of extra credit that was awarded. As a result, participants were rewarded for their participation at varying degrees depending upon what class they were enrolled in. However, I have no reason to believe that the participants were aware of this fact, and I also do not believe that this had any influence on the opinions they provided in the experiment.


A reliability analysis was conducted to determine if constructing a summary variable was appropriate. The summary scale was found to reach an acceptable level of reliability (Cronbach's α = 0.81). In addition, item-total correlations range from .405 to .523, which fall within the recommendation of between .30 and .70 put forth by Kerlinger (1986).


“Anchor appearance” served as a measure of how telegenic the anchor was to the respondent. “Anchor tone of voice” served as a measure of how well the anchor vocally delivered the news. “Anchor level of knowledge” measured how intelligent the respondent thought the anchor was. “Anchor trustworthiness” measured whether the respondent trusted the anchor, and “anchor objectivity” measured whether the respondent thought the anchor was delivering the news in an ideologically neutral manner.


I ran additional models that included control variables for media use patterns (i.e., network preference, hours per week watching TV news, propensity to read the newspaper, propensity to visit online sites, newspaper and/or website used most frequently) as well as three-way interactions between ideology, content, and network attribution. The variables failed to reach significance and did not make a significant contribution to the model. As a result, they were dropped from the final analysis.


CNN content was used in the analysis presented in Figure 1. FNC content could have been used, but given the small coefficients and the high levels of insignificance associated with the content variable and the interaction between content and ideology, no significant substantive difference would have emerged by creating the figure using content presented by the other network.



I would like to thank Jeff Mondak for his guidance in this undertaking. I am also grateful to Bob Jackson, Rod Lewis, Charles Barrilleaux and the editors and anonymous reviewers for their constructive editorial and methodological advice, as well as Karen Halperin for her help in conducting the experiment. In addition, a tremendous amount of gratitude goes to Phil Streetman and WSST-TV 55 in Cordele, GA for their assistance in creating the news broadcast replications used in this study. A previous version of this paper was presented at the 2005 Annual Meeting of the Southern Political Science Association.

Copyright information

© Springer Science+Business Media, LLC 2007