The Bullying Pulpit: The Audience Effects of a Partisan Character-Attacking Speaker

Amy Schumacher-Rutherford and Ashley Muddiman*

University of Kansas


Campus speakers, and the protests against them, have sparked debate in the U.S. about declining support for free speech. Yet, the content of such speeches has largely been ignored. Do audiences want to shut down a speaker because that speaker holds a disagreeable position, or is it the way in which that position is conveyed? That is, does disagreement generally or only difference delivered with animus—in this case, with character attacks—drive audiences toward retaliatory action? To answer, we draw from Kenneth Burke's theory of identification to investigate how audiences react to political rhetoric when they encounter character attacks against the political party with which they affiliate. We propose that the very character attacks a speaker uses to achieve identification with a target audience can also cause disidentification that engenders an oppositional audience poised to act against the speaker – in this case, to restrict the speaker's right to speak. We expect that espousing a different opinion absent character attacks will not have this effect, but we do anticipate differential effects based on the type of character attack. For this, we turn to Burke's approach to framing to determine whether character attacks presenting one's in-group party as foolish (comic frame) rather than traitorous (tragic frame) have distinct effects on the audience. We conduct an online survey experiment of U.S. residents to test whether the two types of attacks, compared to arguments that use identification strategies, decrease support for expressive rights in the context of a college campus speech. Our results indicate that character attacks increase the likelihood that participants attribute malevolence to the outgroup political party, which then decreases their support for a speaker's right to speak. Both comic and tragic attacks lead to the same outcomes. Optimistically, civilly disagreeable speeches that use identification strategies prompt normatively beneficial outcomes, suggesting that not all disagreeable content decreases free speech support. Conversely, character attacks prompt disidentification that leads to retaliatory action. These findings indicate that audiences' free speech support may be more dependent on a speaker's use of character attacks than their issue content.

Keywords: comic frame, identification, Kenneth Burke, tragic frame


In February 2019, administrators at Grand Canyon University, the largest Christian university in the United States, announced they would disallow conservative commentator Ben Shapiro's upcoming speech on their campus (Daily Wire, 2019). This incident occurred in the midst of a heated national conversation, particularly on the political right, about restrictions on free expression within academia that ultimately prompted President Donald Trump to sign an executive order linking federal dollars to universities' enforcement of "free inquiry" (Haberman & Shear, 2019) and at least 17 states to enact free speech laws (Bauer-Wolf, 2019). Indeed, Gallup (Jones, 2018) found that 61 percent of Americans think that university campus climates prevent people from speaking freely. What stands out to us about the GCU incident, however, is the explanation given for the de-platforming of Shapiro. The GCU president deemed him too "cutthroat" and explained, "We really believe that we're so divided as a country right now, extreme positions on the left and extreme positions on the right aren't helping" (Daily Wire, 2019). As these statements show, both style and substance were offered as reasons to nix the planned speech.

Administrators are not alone in fearing the potential consequences of divisive discourse; in their annual civility public opinion poll, marketing firm Weber Shandwick (2019) found that the vast majority of Americans believe that incivility engenders consequences ranging from intolerance and loneliness to harassment and violent behavior. Although there are doubtlessly numerous factors that contribute to why a speaker may or may not be allowed to speak, we focus on whether one of the most frequently-offered concerns—the fear of negative consequences on and/or from the audience—has merit by exploring the effects of a controversial speech on audience members. In this study, we analyze the effects of verbal character attacks in the context of political speech, investigating whether disagreement per se or the nature of the disagreement (e.g. respectfully recognizing a difference in opinion versus demonizing someone for a differing opinion) affects support for the expressive rights of a controversial speaker at a generic American university and protesters responding to them.

To explore this topic, we propose that Kenneth Burke's theory of identification can be used to explain the reactions from members of a group targeted by character attacks. Although he is more widely known as a literary critic and philosopher of language, Overington (1977) argued that sociologists should recognize the contributions and utility of Kenneth Burke's work. Similarly, Branaman (1994) noted that although Burke has influenced American sociology, he has not yet received his due as a social theorist. We hope to demonstrate here the value his ideas can provide applied social theorists generally and character assassination scholars particularly.

Through Burke's work, we argue that a rhetor or text that fails to achieve identification with receivers can, conversely, cause disidentification that constitutes them as an oppositional audience poised for political action. We use a survey experiment to test this phenomenon, expecting that the oppositional subjectivity will lead targets of character attacks to attribute malevolence to the attacking speaker and that speaker's political group (Warner & Villamil, 2017)—that is, we expect that, compared to speakers who use civil identification strategies, those who engage in character attacks will prompt audience members to believe that the attacking political party is deliberately sabotaging the country (an operationalization of malevolence; Warner & Villamil, 2017). This perception of malevolence, then, we anticipate, facilitates action, which, in this context, amounts to decreased support for the expressive rights of the speaker and increased support for people protesting against the speaker. We, thus, expect that when one uses their bullying pulpit literally, it will result in negative outcomes. We also explore whether different types of verbal character attacks have stronger effects on support for expressive rights than others by integrating another of Kenneth Burke's (1937) theories – of comic and tragic framing – with character assassination literature.

Because social theory is a "basic survival skill" practiced by both professionals and lay persons (Lemert, 2018), this study seeks to contribute to conversations by both. By using a socio-theoretically-based experimental design to investigate the mechanisms and outcomes of current, real-world situations relating to adversarial politics, we hope to exemplify a quantitative approach to social theory and provide relevant and actionable conclusions for practitioners and consumers of political media. Through this study, we also aim to contribute to character assassination literature in three ways—by examining attacks against a group rather than individual, by merging social theory and quantitative methods, and by examining the effects of character attacks, which has been insufficiently investigated (Shiraev, 2014).

Identification and Creation of an Oppositional Audience

Building identification with an audience is an essential part of persuasion generally and political speech specifically. From the Burkean perspective (1969), identification involves speakers attempting to transcend differences and become a collectivity with the audience. He explained identification with the following example: "A is not identical with his colleague, B. But insofar as their interests are joined, A is identified with B. Or he may identify himself with B even when their interests are not joined, if he assumes that they are, or is persuaded to believe so…He is both joined and separate, at once a distinct substance and consubstantial with another" (1969, pp. 20-21). In other words, humans are autonomous but have an inherent need to identify with others, giving speakers the opportunity to build consubstantiality with their audiences—regardless of the veracity of the perceived sameness—and then successfully persuade them. For example, Lupia (2013) found that it is perceived interest commonality—in other words, identification—that drives audience compliance with scientific recommendations in a politicized environment. From this perspective, audience members have agency too—they can choose to accept a rhetor's invitation (Foss & Griffin, 1995) to consubstantiality or reject it.

Other scholars have similarly expounded on the concept of identification. Edwin Black (1970), for instance, proposed that any given text offers the audience an impression of who the speaker is—a perceived persona, which he termed the first persona—and an implied audience—a textual construction concerning the ideal or target audience, which he called the second persona. This understanding of first and second personae contributes to the theory of identification by recognizing that both a speaker and the audience exist as separate personae crafted through the speaker's text. The components of a text that rhetorically construct the ideal audience serve as an invitation, one "not simply to believe something, but to be something. We are solicited by the discourse to fulfill its blandishments with our very selves" (Black, 1970, p. 119). This process illustrates what Burke meant by the consubstantiality that occurs with successful identification—the text can construct an audience that is drawn toward and becomes one with the speaker. Further, Charland (1987) borrowed Althusser's concept of interpellation to describe a similar phenomenon, arguing that language can, through identification, constitute—or call into being—subjects who are induced to support a speaker's position because doing so fulfills the expectations of their subjectivity. To demonstrate this theory of constitutive rhetoric, Charland used the example of the independence movement in Quebec and its construction of a new, collectivized subjectivity—peuple québécois—which, through identification, interpellated Quebecois individuals as political subjects whose support of sovereignty was inherent to their new subject-position. In sum, speakers can use identification strategies in their addresses to constitute an audience that is poised to support them both ideologically and materially.

Often, scholars studying identification focus on the particular audience with which the speaker—usually a political figure—is attempting to identify. Connaughton and Jarvis (2004), for instance, examined how political parties and presidential candidates used political advertisements to build identification with Latinx voters, Benoit (2000) explored how presidential candidates Bill Clinton and Bob Dole built identification with voters in 1992 through their television advertisements, and Bimber (2014) describes presidential candidate Barack Obama's novel and effective use of personalized political communication. Yet, there is a secondary audience important to political messages: the audience not originally meant to see a message. Even with concerns about microtargeting (Kim et al., 2018) and ideological filter bubbles (Pariser, 2011), members of the public report exposure to political disagreement, especially in digital spaces (Barnidge, 2017). Anecdotally, when campus speeches are cancelled or protested, or if they are allowed a platform but contain controversial elements, they are often picked up by news media, which share the messages of the speakers with a mass audience. Thus, it is necessary to understand how efforts to solidify identification with an intended audience can simultaneously result in disidentification among unintended audience members outside the target group, who may then want to take retribution against the speaker.

Speakers thus have two choices to make when building identification with their audiences: First, do they build identification with their intended audience in a way that also invites identification with the unintended audience, or do they build identification with the intended audience in a way that amplifies divisions with the unintended audience? Second, if they choose the latter, do they treat the unintended audience as incompetent or as villains? We address each of these in turn.

Identify or Divide?

On one hand, some speakers choose to use language to build identification with both the intended audience and the unintended audience. A Democratic speaker could acknowledge differences in policy stances from the Republican Party while also mentioning ways that the party members have similar, well-intentioned goals. For example, during the 2008 U.S. presidential campaign, Republican candidate John McCain said of his Democratic opponent Barack Obama, "He's a decent family man and citizen that I just happen to have disagreements with on fundamental issues" (Martin, 2008). More recently, in the 2020 U.S. Democratic primary election campaign, Pete Buttigieg said of "future former Republicans," "I'm running on the values that make me a Democrat, but there is room for a lot of people" (Schneider, 2019). These rhetorical choices invite people outside of the main intended audience to create identification with the speaker despite differences in party identification and political opinion.

A speaker can, alternatively, in an attempt to build identification with the intended audience, amplify divisions with an unintended audience—creating an opportunity for the unintended audience to coalesce into an oppositional audience through a process of disidentification. Wander (1984) argued that, by constructing an ideal audience through their text, speakers also construct an opposite; they affirm desirable characteristics and invite one audience to both inhabit and act upon them, but they simultaneously assert which behaviors and beings ought to be avoided. This observation echoes Burke (1969), who noted that "identification is compensatory with division" (p. 22); he explained that without division, there would be no need for identification. In fact, although the most elementary way to build identification is to state one's similarities with the target group (Cheney, 1983), an often more successful way of achieving identification within a group is to "join forces against a common enemy" (Burke, 1972, p. 28). For instance, in contrast to McCain's approach to Obama mentioned above, President Donald Trump often emphasizes division with the opposing party, for instance by claiming "the radical left Democrats in Washington are trying to burn it all down" (Trump, 2020). Thus, a speaker from one political party can attack the other political party in a speech, increasing identification with audience members from the speaker's own party by amplifying divisions with the oppositional, unintended audience; in doing so, however, they may spur action among the oppositional audience through disidentification.

Disidentification refers to "the active rejection and distancing of a particular group" (Verkuyten & Yildiz, 2007, p. 1456), but the concept is empirically distinct from a mere low sense of belonging or dislike of a group. In fact, Elsbach and Bhattacharya (2001) found that disidentification goes beyond psychological distance and can lead individuals to take action against the entity from which they disidentify; in their case, people who disidentified with the National Rifle Association (NRA)—thereby perceiving themselves as oppositional to that organization—were motivated to speak out and/or volunteer against the NRA. Disidentification in response to divisive rhetoric can therefore be understood as evidence of a constituted oppositional public; we seek to understand if a particular form of divisiveness—character attacks—facilitates the disidentification process.

Investigating the effects of divisive language can inform character assassination scholarship as well. Character assassination involves an intentional attempt to harm a target's reputation (Icks & Shiraev, 2014) and can be accomplished through a variety of methods of character attacks including, most notably for this paper, demonization through name-calling that reduces the target to negative traits (Shiraev, 2014) and fearmongering that presents the target as an enemy or threat (Icks et al., 2020). Because "all national politics [in the U.S.] take place in a context of permanent, professionally-managed, and adversarial campaigning" (Heclo, 2000, p. 30), American political actors are engaged in perpetual efforts to undermine their oppositional party, which has yielded all-time high levels of animosity toward one's opposing party (Pew Research Center, 2017). In fact, echoing Burke's words on achieving identification through uniting against an enemy, negative partisanship—the practice of identifying against a party rather than for the other—has developed and expanded in the United States over the past few decades (Abramowitz & Webster, 2018).

This contentious political context provides favorable conditions for character assassination to occur, for if attackers successfully undermine the character of their opponents, they are more likely to achieve electoral or legislative victory (Shiraev, 2014). Indeed, character attacks are used to achieve identification with and beget action from the intended audience by labelling the opposition with derogatory terms, and name-calling has been a consistent feature of American politics since the nation's inception (Smart & Shiraev, 2014). However, as Shiraev and Makhovskaya (2020) explain, character attacks not only have direct psychological effects on the specified target, but they can also inflict indirect psychological harm on individuals who identify with the victim of character assassination.

Further, although most character assassination research has focused on the destruction of individual reputations, groups too can possess character that comes under attack (Icks et al., 2020). For instance, scholars have argued that corporations (Coombs & Holladay, 2020) and countries (Aspriadis, Takas, & Athanassios, 2020) can be targets of character assassination. We extend this logic to the assassination of the character of a political party, not just individual politicians, and investigate the effects of viewing character attacks against one's political party. Many scholars have argued that the influence of political parties in the United States is on the decline (Everson, 1982), yet there is other evidence that, especially since the great American partisan sort (Levendusky, 2009) and digital advances, parties are experiencing a renewal phase (Chadwick & Stromer-Galley, 2016). What is most important for the present project, though, is that party identification remains a source of social identity (Greene, 2004), even among individuals who claim to be independent (Greene, 1999), and serves as an umbrella under which numerous traits and preferences align (Johnston, Lavine, & Federico, 2017).

As Icks et al. (2020) explained, while character refers to the actual traits a person has, character assassination involves attacks on reputation, or the perception of the traits one has. American partisans tend to have both actual trait differences—Hibbing, Smith, and Alford (2013), for example, note that numerous studies over 70 years have linked personality dimensions to preferences that reliably sort liberals and conservatives—and perceived ones that have become stereotypical in the present culture wars (e.g. "latte liberals") (DellaPosta, Shi, & Macy, 2015). We thus contend that although parties are more of an amalgamation of networks and shared values than monolithic entities (Chadwick & Stromer-Galley, 2016), they not only have character but also reputations ripe for character assassination. In this project, then, we test whether viewing a political speech containing character attacks against one's own political party can lead to disidentification, constituting an oppositional audience poised to act against the attacker. The severity of the oppositional reaction, however, should be based on the type of attack deployed.

Incompetent or Traitorous?

Kenneth Burke elucidated two types of character attack that speakers can use once the decision has been made to amplify division. Burke (1937) contended that the particular negative labels people choose to use against others cue differential responses. He argued that selected labels suggest not just who you should be for or against, but also "how you shall be for or against [an opponent]. Call a man a villain, and you have the choice of either attacking or cringing. Call him mistaken, and you invite yourself to attempt setting him right" (p. 4). In other words, the language one uses to describe those with whom they disagree constrains the means of response available—a villain is to be terminated, not considered an equal worthy of collaboration or compromise. Burke (1937) categorized the fool/unintelligent label as the comic frame, while the villain/traitor approach constitutes the tragic frame. These frames align well with two types of character attacks outlined by Icks et al. (2020)—name-calling with the comic frame and fearmongering with the tragic frame—and can thus provide insight into the effects utilizing these tactics can have.

The notion of a tragic frame also aligns with newer comparative political theory work demonstrating that authoritarian leaders tend to use character attacks that delegitimize other politicians as enemies that need to be destroyed rather than political opponents with different viewpoints (Levitsky & Ziblatt, 2018). The devastating effects of tragic framing have been observed throughout history, with the construction of enemy images that dehumanize an other (Keen, 1991) leading to support for—if not participation in—individual and group-level violence (Beck, 1999), including genocide (Gendron, 2012). Conversely, Burke (1937) argued that the comic frame uniquely enables reconciliation, asserting that "the progress of humane enlightenment can go no further than in picturing people not as vicious, but as mistaken" (p. 41). Similarly, George Orwell (1968) asserted the import of language choices, explaining that while using slovenly language can beget foolishness, clear language can contribute to political regeneration. Indeed, Cram (2017) argued that it was President Obama's use of the comic frame in his foreign policy rhetoric toward Iran—a shift from the tragic frame that dominated President Bush's post-9/11 foreign policy rhetoric—that enabled a peaceful resolution to the nuclear standoff at that time.

Various strategies of identification and division have been identified by previous researchers, but the current project extends the literature by measuring the effects of such language choices, particularly the responses of unintended audiences when a speaker attempts to identify with them despite their differences compared to the responses when a speaker actively negates them in discourse using comic-frame or tragic-frame character attacks. Identification scholarship has typically focused on the persuasive effects of building identification and division among audiences. Burke (1969), for instance, considered identification a necessary precondition for persuasion, and political scholars have examined rhetorical identification-building techniques in election contexts where messages are created to persuade the audience to vote for a candidate or party (Benoit, 2000; Connaughton & Jarvis, 2004). While we focus on speeches that include persuasive arguments, we extend identification research by exploring the effects of identification and division efforts on attributions of malevolence and support for expressive rights.

Attributions of Malevolence

As we investigate the relationship between viewing character attacks against one's political party and supporting retaliatory actions, we have reason to expect the path to be indirect, and that malevolence serves as a mediator. Previous research (Munro, Weih, & Tsai, 2010) has found that political rhetoric from an oppositional party is not enough to induce strong negative feelings about the party, but the presence of perceived malevolence does have that effect. And, as we argue below, measuring malevolence is a way to assess whether disidentification has occurred, and we do not expect individuals who did not disidentify to support the retaliatory responses.

In psychological circles, malevolence typically refers to the three negative personality traits comprising the Dark Triad (Furnham et al., 2014). However, in this project, we rely instead on the following socially-focused definition: malevolence as the belief that one's political opposition is illegitimate and has ulterior motives (Warner & Villamil, 2017). This approach to malevolence is centered on a person's belief about why their opposition is behaving in a given way (see, for instance, Pettigrew, 1979). A Democrat may, for example, believe that although she disagrees with the Republican Party's issue positions, Republicans are trying to do what is best for the country (low level of perceived malevolence) or that Republicans are out to ruin the country (high level of perceived malevolence). Although malevolence attributed toward one's in-group may be little more than relatively healthy skepticism (Jennings et al., 2018), it becomes more harmful when aimed at an opposing group. Valentine, Hanson, and Fleischman (2019), for instance, found that perceived malevolence leads to greater acceptance of unethical treatment for the opposition party.

Given identification theory, as outlined above, we expect that if a speaker's use of character attacks against the opposing political party leads to disidentification among members of the attacked group, this process will manifest in attributions of malevolence to the political party of the attacker. Although malevolence appears to have not yet been used in disidentification research, the related concept of cynicism has (Roderick, 2014). If individuals perceive a group as so diametrically opposed to themselves that they actively disidentify and constitute an oppositional audience, we expect them to not just be distrustful (cynical) about the offending group, but to also perceive them as a malevolent entity. Indeed, we propose that malevolence can serve as one measure of disidentification because it demonstrates that individuals are not only psychologically distant from the attacking group, but are so removed that they view them as malicious.

That said, if a speaker has made an effort to build identification with the oppositional party (e.g. a Republican speaker acknowledges the humanity of Democrats), we expect that oppositional audience members exposed to the argument will feel more identified (Burke, 1969) with that speaker and, thus, less be likely to perceive the motives of the speaker's political party as malevolent. Therefore, although they may disagree with the speaker and the speaker's party, they will likely not disidentify with them or seek retaliatory action. If, however, a speaker is amplifying division with the outgroup in order to build identification with the intended audience (e.g. a Republican speaker attacks the character of the Democratic Party and its members), we expect an oppositional audience member to actively disidentify with the speaker's party and attribute more malevolence to that party. Further, because the tragic frame questions the very legitimacy of the oppositional audience, we predict that the intensity of the effect will be heightened when a speaker uses tragic-frame character attacks rather than comic-frame character attacks (Burke, 1937).


H1: When received by an unintended audience, speeches that emphasize division through character attacks against a political party will prompt more attributions of malevolence to the speaker's political party than speeches that use civil identification strategies.

H2: When received by an unintended audience, speeches that use tragic-frame attacks will prompt more attributions of malevolence to the speaker's political party than speeches that use comic-frame attacks.


Finally, and most importantly, we expect that if the attacks incite feelings of malevolence toward the speaker and/or the speaker's party, the disidentified audience will be oriented to act in retaliation. Previous research has found that when members of one group view another group as harmful—in other words, malevolent—they are more likely to support removing that group from their environment if not actively bring the group harm (Halperin, 2008). This study investigates three outlets of retribution: a desire to restrict the speaker's expressive rights, the wish to restrict free speech generally, and support for the expressive rights of protesters conveying their opposition to the rhetor.

Research regarding support for free speech has primarily been conducted by think tanks and polling groups. The majority of Americans tend to support allowing offensive speech even if they personally disagree with the content; for instance, The Pew Research Center (Wike, 2016) found that 67 percent of Americans think people should be allowed to make public statements minority groups find offensive and the CATO Institute (Ekins, 2017) found that 79 percent agree it is morally unacceptable to engage in hate speech against racial or religious groups. Yet, there is evidence of a gap between support for free speech rights in the abstract and support for more concrete, real-world examples (Knight Foundation, 2018). Accordingly, when we measure support for expressive rights as an outcome variable, we use separate measures to assess support for free speech generally—e.g. if someone should be able to say something offensive to minority groups in public—and more specifically for the speaker and people protesting the speaker, asking if the speaker whose speech they read and the protesters about whom they read a news article should be allowed to speak in public places.

Malevolence attributions have been shown to predict outcomes as extreme as support for political violence (Warner & Villamil, 2017). As described above, we view malevolence as evidence of the process of disidentification, which constitutes individuals as an oppositional audience; once they are disidentified, to act against the attacking speaker and the speaker's party is inherent to their new subjectivity as an oppositional audience. When someone is the target of character attacks, there are several options of how to respond available to them, including counterattacks, which can be particularly effective when deployed through surrogates (Minielli, 2020). We similarly predict that attributions of malevolence will drive a response of retribution—that is, a counterattack of sorts—toward the speaker. As an audience member reports a stronger belief that the speaker's political party has less-than-pure motives, we anticipate decreased support for the speaker's rights specifically and free speech rights generally as these are the retaliatory options presented to them. Additionally, we expect perceived malevolence to beget increased support for protesters standing up against the speaker as those demonstrators can be seen as fellow oppositional audience members and surrogates for the actions they would like to take but cannot.

H3: As outgroup malevolence increases, (a) support for the speaker's expressive rights decreases, (b) support for protesters' expressive rights increases, and (c) support for free speech in general decreases.

In sum, we offer an indirect effects model predicting that the approach a speaker takes to an oppositional audience (civil identification, comic-frame character attacks, tragic-frame character attacks) will affect how the outgroup audience members attribute malevolence to the speaker's party, which then affects support for the expressive rights of the speaker, for the expressive rights of the protesters who oppose the speaker, and for free speech generally (see Figure 1). We further expect the overall indirect effect to vary when a speaker uses tragic-frame character attacks rather than when the speaker uses comic-frame character attacks. In sum, we predict:

H4: When speeches that include character attacks are received by the members of the attacked group, compared to speeches that use civil identification strategies, will indirectly (a) decrease support for the speaker's expressive rights, (b) increase support for protesters' expressive rights, and (c) decrease support for free speech in general through attributions of malevolence toward the speaker's political party.

H5: The indirect effect relationships will be stronger when the speech includes the tragic-frame attacks rather than to the comic-frame attacks.


Graphic of Predicted indirect effect model

Figure 1 Predicted indirect effect model. + indicates a predicted positive path and – indicates a predicted negative path



Given that our primary question concerns the measurable effects speeches utilizing character attacks have on audience members, the most appropriate method to answer it is an experimental design. We recognize that quantitative methods are not often used in character assassination or Burkean research; however, utilizing them here enables us to answer different questions than previous work—the consequences these theories have in practice. And, using an experimental study—rather than simply an observational one—empowers us to determine cause and effect (Dean, Voss, & Draguljic, 2017); specifically, the effects character attacks have on audiences (if they disidentify and/or if they seek retaliation) and whether comic and tragically-framed attacks cause different responses, as Burke argued they should. We thus conducted an online, between-group, survey-based experiment with three conditions, exposing participants to a policy speech from their political outgroup containing either civil identification or character attacks (either comic or tragic). We then measured the messages' effects on perceived malevolence and our outcome variables relating to support for expression.


We recruited U.S. participants using Amazon's Mechanical Turk (Mturk) platform in May 2018. Because we wanted our results to be applicable beyond a particular campus and because, given our current media environment, controversial speeches tend to circulate well beyond the location and context in which they are given, we chose to use general public participants without a campus affiliation. Further, although Mturk samples are not nationally representative, they are closer than university subject pools (Paolacci et al., 2010). Because we recruited participants using Mturk's qualification of "Political Affiliation," our sample was politically diverse; of 371 respondents, 196 identified to Mturk as liberals and 195 identified to Mturk as conservatives. Those affiliations largely remained consistent with participants' self-reported party affiliation (which was measured after exposure to the stimuli) with slightly more Democratic individuals (Range: 1 = strong Republican through 5 = strong Democrat; M = 3.21; SD = 1.48). The average age of our participants was 44.46 years old (SD = 13.86), 48 percent were female, and they had an average of 15.39 years of schooling (SD = 1.80). Additionally, 48 percent of respondents reported themselves as White/Caucasian, 5 percent as Black/African-American, 4 percent as Asian/Pacific Islander, and 4 percent a different race.


We first presented participants with the transcript of a political speech about criminal justice reform we said was given by an unnamed speaker on a college campus (the generic "State College"). They then reported their level of perceived malevolence toward the speaker's political party as well as their level of support for the speaker's right to speak. Next, we showed participants a news article about a violent protest that occurred in response to the speech, and, afterward, measured their support for the expressive rights of the protesters and for free speech generally. Finally, participants provided information regarding their demographics and partisanship and were paid for their participation.


All participants read a persuasive policy speech about criminal justice reform. This topic was selected because it is both a high-profile political topic – it has been championed by public figures ranging from Senior White House Advisor Jared Kushner (Min Kim & Gearan, 2018) to Kim and Kayne West (Sullum, 2018)—and yet is a relatively low-level priority among the American public, not even making the list of potential top policy issues in Pew Research Center (Pew Research Center, 2019) or Associated Press (Riccardi & Fingerhut, 2019) annual surveys. Further, there is a great deal of common ground on the issue; the criminal justice reform bill that President Trump signed into law seven months after we conducted our experiment had broad bipartisan support and passed by large margins in both chambers (Rodrigo, 2018). Thus, because criminal justice reform has some public salience but little comparatively public division, the topic enabled us to more acutely home in on participants' responses to the verbal attacks per se rather than the policy being advocated.

There were six different speech stimuli; participants received one given by a member of their outgroup party (i.e. a conservative received a speech from the Democrat). The Republican and Democratic speeches were nearly identical in both structure and length. They differed only in which party was named, which solutions the speaker proposed (e.g. eliminating mandatory minimum sentencing in the Democratic speech and expanding the use of private prisons in the Republican speech), and the verbiage of the character attacks.

Participants were randomly assigned to view a speech from one of the three speech conditions within their outgroup. There were 10 places within the 800-word speeches where we manipulated the civil/comic/tragic wording. In the civil identification condition, we included sentences where the speaker emphasized respect, common goals, and other strategies of civil disagreement. For instance, the speaker mentioned Republicans and Democrats coming together for a solution (e.g. "we're going to need a bipartisan solution to best combat this flawed status quo" and "we need to put aside our differences and solve this problem"), respected the intentions of the other party (e.g. "even Democrats/Republicans admit there are issues with such a large prison population and many have good intentions in their approach to solving this problem"), and recognized a common set of facts and goals (e.g. "taxpayers are paying about $80 billion a year to keep so many people in prison. We can all agree that this is way too high" and "both Republicans and Democrats know how incarceration and drug abuse can ruin a promising life, can rip apart a family, can weigh down an entire community").

To maximize authenticity, most of the insults used in the character attack conditions were minimally-edited jabs taken from public social media posts. The comic-frame attacks focused on the other party as incompetent. We used many of the insults interchangeably between the parties (e.g. "weak Republican/Democratic Party", "idiot Republicans/Democrats"), but also employed real-world, partisan-specific attacks (e.g. "spineless Republicans" versus "snowflake Democrats" and "wingnut GOPers" versus "libtards"). These attacks purposefully did not condemn the motives of the other party, but rather their intelligence and aptitude. The tragic frame condition presented a starkly different type of character attack, with the other party being framed as a conniving, bad-faith enemy (e.g. "villainous Republicans/Democrats," "devious Republicans/Democrats," and "fascist conservatives" versus "socialist liberals") that does not abide by mutual facts (e.g. "corrupt conservatives/liberals won't admit that's a problem") and is a direct threat to the other party (e.g. "Republicans/Democrats, content to destroy all we hold dear, have fought against this policy, and they're winning" and "Republicans/Democrats, by willfully ignoring the evidence are traitors to the American people").

Measure of Attributions of Malevolence

Finally, five items comprised our malevolence attributions measure based on Warner and Villamil (2017). Using a five-point Likert scale, participants reported their level of agreement with the following statements: I worry that the Republicans/Democrats are deliberately trying to hurt America, the Republicans/Democrats are knowingly sabotaging the country, the Republicans/Democrats don't care about America, I believe the Republicans/Democrats genuinely want what is best for America, and I trust the Republicans/Democrats to do what they think is best for America (M = 3.20; SD = 1.131; Cronbach's a = .90). Like Warner, et al. (2019), we use this measure of attributions of malevolence as a mediator leading to a political outcome—in our case, support for free speech generally, of the speaker, and of protesters.

Measures of Expression Support

The first measure of expression support we examined was support for the speaker's right to speak, which participants completed shortly after reading the speech. Our support for speaker's free speech measure had three items with a five-point Likert scale of agreement: this speaker has a right to speak, this speaker should not be allowed to speak in public places, and this speaker has a right to be heard (M = 4.37; SD = .86; Cronbach's a = .78).

Next, we measured support for protesters' expressive rights. After participants read the speech and reported their support for the speaker's right to speak, participants read a short news article modified from a Los Angeles Times story about a violent protest that broke out after an on-campus speech. No information about the issue positions of the protesters was mentioned, so all participants, no matter whether they read a speech attributed to a Democrat or a Republican, could read the same article about the protest. The article emphasized the violence and police presence, and violation of the status quo and did not address the protesters' concerns, all typical characteristics of protest media coverage (McLeod & Detenber, 1999). To test whether participants supported the expressive rights of the protesters, we used McLeod and Detenber's (1999) three-point Likert scale: these protesters have a right to protest, these protesters should not be allowed to protest in public spaces, and these protesters have the right to be heard (M = 3.78; SD = 1.01; Cronbach's a = .79).

We operationalized our third measure of support for expression—support for general free speech—by adapting the five-item measure used by the Pew Research Center (Wike & Simmons, 2015). Respondents either said, yes, people should be able to say the items publicly or, no, the government should be able to prevent people from saying these things. The five topics to which they responded were: statements that criticize the government's policies, are offensive to minority groups, are offensive to your religion or beliefs, call for violent protests, and that are sexually explicit (Range: 0 to 5; M = 2.95; SD = 1.63; Cronbach's a = .78).


As previously explained, we expected the comic/tragic attacks to prompt attributions of malevolence, which would then lead to lower levels of support for the speaker's expressive rights and higher support of the protesters' expressive rights. Thus, to test our hypotheses, we ran indirect effect analyses using Haye's (2018) SPSS macro, Model 4. Speech content was our predictor variable; we set civil identification speeches as the reference group, meaning that a positive result would denote that the comic- or tragic-frame attack increased the outcome variable (e.g. support for expressive rights) through the mediator, malevolence attributions. Conversely, a negative coefficient would mean the comic- or tragic-frame attack decreased support for the outcome variable (e.g. expressive rights). Following Hayes' (2018), we included control variables – including age, sex, conflict avoidance, and partisan strength (Mutz, 2015; Warner & Villamil, 2017) – that previous research indicated could be related to the measured mediating and outcome variables. We tested the direct effects (i.e. H1 through 3) using the OLS regression models prepared by PROCESS. The significance of the indirect effects was determined using confidence intervals, as recommended by Hayes (2018); thus, for the indirect effects, we report the 95% confidence intervals bias corrected by 5,000 bootstrapped samples.


H1 predicted that speeches emphasizing division against a political party would prompt more attributions of malevolence toward the speaker's party than speeches using identification strategies. This hypotheses was supported. Results of the significant OLS linear regression model predicting attributions of malevolence [R2 = .05, F(6, 345) = 3.05, p < .01] demonstrate that both comic-frame (b = 0.33, SE = 0.14, p < .05) and tragic-frame (b = 0.30, SE = 0.15, p < .05) attacks lead to higher levels of malevolence attributions than speeches that incorporated identification strategies. These results also test H2, which predicted that tragic-frame attacks would prompt more attributions of malevolence than comic-frame attacks. The coefficients for the comparisons between identification strategies and comic-frame attacks, as well as between identification strategies and tragic-frame attacks were nearly identical, suggesting that these different types of character attacks did not prompt different amounts of malevolence toward the speaker's party. H2 was not supported.

Next, we examined H3, which predicted that increases in attributions of malevolence predicted (a) decreased support for the speaker's expressive rights, (b) increased support for protesters' rights, and (c) decreased support for free speech rights generally. H3a was supported by a significant OLS regression model [R2 = .27, F(7, 347) = 3.96, p < .001]: increasing attributions of malevolence were significantly related to decreasing levels of support for the speaker's expressive rights (b = -0.16, SE = 0.04, p = .001). Although the regression models predicting support for protesters' rights [R2 = .06, F(7, 344) = 3.19, p < .01] and support for free speech generally [R2 = .33, F(7, 347) = 6.14, p < .001] were significant, the attributions of malevolence variable was not a significant predictor of either outcome (support for protesters' rights: b = 0.04, SE = 0.05, p = .36; support for free speech rights generally: b = 0.01, SE = 0.07, p = .85). Thus, there was no support for the relationships predicted by H3b and H3c.

We now turn our attention to the indirect effects models. First, we tested H4a, which predicted that exposure to a speech that used division and character attacks would decrease support for the speaker's expressive rights through attributions of malevolence compared to a speech that emphasized identification strategies. This hypothesis was supported (see Table 1, Model Y1). Compared to speeches that incorporated identification strategies, both the comic- and the tragic-frame attacks significantly decreased support for the speaker's right to speak.

  Effect Coefficient Bootstrapped S.E. 95% Confidence Interval
Message Frame (X) à Malevolence (M) à Support for Speaker's Expressive Rights (Y1)
Civil Identification v. Comic-Frame -0.06 0.03 (-.1303, -.0097)
Civil Identification v. Tragic-Frame -0.06 0.03 (-.1227, -.0028)
Message Frame (X) à Malevolence (M) à Support for Protester's Expressive Rights (Y2)
Civil Identification v. Comic-Frame 0.01 0.02 (-.0211, .0595)
Civil Identification v. Tragic-Frame 0.01 0.02 (-.0217, .0555)
Message Frame (X) à Malevolence (M) à Support for General Free Speech Rights (Y3)
Civil Identification v. Comic-Frame -0.005 0.02 (-.0425, .0297)
Civil Identification v. Tragic-Frame -0.004 0.02 (-.0385, .0284)

Table 1 Indirect Effects Predicting Support for Expressive Rights

Note: The indirect effect coefficients are partially standardized.


H4b and H4c were, however, not supported (See Table 1, Model Y2 and Y3). H4b predicted that exposure to a speech with character attacks would increase support for protesters' expressive rights indirectly through attributions of outgroup malevolence, but neither comic-level nor tragic-level character attacks significantly affected support for the protesters' expressive rights compared to speeches that incorporated politeness. H4c predicted that a speech with character attacks would decrease support for general free speech indirectly through attributions of outgroup malevolence. As indicated in Table 1 Model Y3, this indirect effect was not significant.

Finally, we tested whether the strength of the indirect effects differed based on whether a participant was exposed to a speech with comic-frame or tragic-frame character attacks (H5). This hypothesis that the effects would be stronger in the tragic condition was not supported. In all models, the partially standardized indirect effect coefficients for the comic-frame and tragic-frame character attacks were, with rounding, often identical (Table 1, Model Y1: B = -0.06; Model Y2: B = 0.01; Model Y3: Comic-Frame B = -0.005, Tragic-Frame B = -0.004). This suggests that the two paths had similar strengths no matter the outcome variable of interest.


This study provides evidence that speeches using strategies of division with an outgroup audience, compared to those that use strategies of identification with that audience, increased attributions of malevolence to the speaker's political party, which, in turn, decreased support for an offending speaker's expressive rights. However, the character attacks did not affect support for protesters' expressive rights or support for free speech more generally, nor did we find that the type of character attack matters in the strength of these relationships.

Our findings suggest that some audiences recognize, and appreciate, when a speaker uses identification strategies (Lupia, 2013) when speaking of the political opposition. When a speaker from a participant's opposing political party made an effort to build consubstantiality with that opposing political party (Burke, 1969), that participant was less likely to think the speaker's political party was undermining the country and, in turn, less likely to desire limiting the expressive rights of the speaker. Alternatively, character attacks of any type prompted a desire for retaliation against the speaker, which we argue stemmed from the conditions inherent to disidentified individuals' (Verkuyten & Yildiz, 2007) new subjectivity (Charland, 1987) as an oppositional audience. Notably, the policy positions in the speeches did not differ—only the way that the speaker treated the unintended audience in the text—which suggests that character attacks and divisive rhetorical strategies, not policy differences, led to the oppositional audience's desire to counterattack (Minielli, 2020). In other words, it was character attacks, not disagreement, that—through disidentification/perceived malevolence—led individuals to want to restrict an offending speaker's right to speak in public. Previous research has found that affective polarization is so extensive, hostility toward one's political outgroup can be automatically cued (Iyengar & Westwood, 2015). Our results, however, indicate that partisans are more willing to support the expressive rights of political adversaries if they disagree while seeking to build identification across the political divide. Accordingly, partisans can advance their distinct policies without alienating those who disagree with them.

Importantly, the decrease in support for free expression only applied to the specific speaker—not to free speech support more broadly or to the support of free expression for violent protesters, confirming previous evidence (Knight Foundation, 2018) of higher support for free speech in general than in specific cases. It appears that the fears expressed by editorial journalists covering controversial public speakers on college campuses (Daily Wire, 2019) and lawmakers scrabbling to pass "free speech" laws (Bauer-Wolf, 2019) may be overblown. The findings in this study provide evidence that when individuals encounter a text in which they are the target of character attacks, they may disidentify with the attacking group strongly enough to constitute an oppositional public. The malevolence they attribute to the attacking rhetor means they find the rhetor dangerous to the country and are thus positioned to counterattack as embodiment of their oppositional subjectivity; yet, the retaliatory action is limited to the offending party, the speaker. Audiences do not seem to desire a shutdown of free speech; instead, they respond negatively to specific speakers who highlight group divisions and add character attacks to arguments that do not alone beget the desire to de-platform. These results should inform scholars' and practitioners' approach to the "free speech crisis."

Worryingly, our results indicated no difference between the comic and tragic character attacks, meaning character attacks per se, not the type of attack, led to the support for restricting the speaker's expressive rights. Burke (1937) argued that the labels one applies guide their behavior and that the comic frame but not the tragic frame enables reconciliation. We therefore find it normatively problematic that calling someone an idiot and saying that person is trying to destroy America render the same results, for this finding implies that any sort of insult can be enough to trigger retaliatory impulses from the target, as limited as they may be. Additionally, these results suggest that, despite the potential for democratically-damaging outcomes (Burke, 1937; Levitsky & Ziblatt, 2018), participants did not respond more stringently to tragic-frame attacks compared to comic-frame attacks.

Despite our normatively optimistic finding that identification efforts can dampen retributive effects of political disagreement, it is important to consider that character assassination attempts are often effective for political candidates looking to win (Shiraev, 2014) and digital news organizations hoping to generate audience engagement with user-generated content (Muddiman & Stroud, 2017). Thus, using character attacks as a method of achieving identification with a political actor's intended audience remains attractive regardless of the consequences for either the targets of the attacks or the democratic system more broadly. Therefore, seeking ways to better motivate political leaders to build identification among political groups is a worthy endeavor necessitating future research.

Our use of quantitative methods to approach character assignation research offered us benefits but also limitations. Through our project, we were able to determine the measurable effects of character attacks on audience members, a feat that non-quantitative approaches are able to do with less certainty. Further, because we used an experimental study, the only type that enables researchers to draw conclusions about cause and effect (Dean, Voss, & Draguljic, 2017), we were uniquely able to make those determinations. We hope we have thus demonstrated the utility of quantitative approaches to character assassination and social theory research. On the other hand, this study was also limited by its scope. Focusing on one experiment and one topic enabled us to identify the processes underpinning the responses of character attack targets, but it also restricted us to investigating one instance of character attacks rather than a broader campaign of character assassination. It was also beyond the scope of this project to test the differential effects of counter-attitudinal speeches versus a speech with which the audience agrees, but that would be a useful inquiry for future research to address. Yet, this study complements the more typical rhetorical (e.g. Gilbert, 2020) and historical (e.g. Egorova & Egorova, 2020) approaches to character assassination research by helping to demonstrate potential effects of character attacks on the audiences that the speakers are attacking.

Additionally, evidence of partisan character attacks mirroring the ones used in this study abound in American political discourse. Given that individuals select their political media exposure based on their political beliefs (Stroud, 2008), and partisan media outlets tend to provide extreme exemplars of the oppositional party (Yang et al., 2016), if partisans view continually exaggerated attacks on their ingroup from the outgroup, we would predict the results we found to be strengthened if not exacerbated. Levendusky (2013), for instance, found that watching partisan news from one's political ingroup led individuals to view the other party as more negative, less trustworthy, and less supportive of bipartisanship. Meanwhile, Levendusky and Malhotra (2016) found that reading a news article about polarization alone was enough to increase dislike of the opposing party. Future work should explore both the long-term effects of attempted party-based character assassination, the effects of cross-cutting exposure to civil versus attacking partisan coverage, and whether exposure to character attacks strengthens the degree to which one identifies with their political party.

Relatedly, previous research has found evidence of partisan differences in responses to the vilification of one's party. Wagner, Mitchell, & Theiss-Morse (2011) found that when Republicans see their party vilified, they are much more negative toward Democrats, but Democrats' views of Republicans do not meaningfully change when they are confronted with vilification of their party. Further, the ways in which the partisans in the study were motivated to act in response to vilification differed, with Democrats more likely to engage with friends about the issue and be more supportive of governmental compromise while Republicans are more likely to donate money to their party and be less supportive of governmental compromise. Thus, we anticipated people may respond differently to the stimuli based on their political preference. In this study, however, we did not find significant partisan differences on any of our key variables. Researchers should continue to explore partisan discrepancies in response to character assassination tactics. Similarly, it would be interesting to determine if particular personality traits such as narcissism affect responses to character attacks.

Finally, in his seminal work on the topic, Aristotle distinguished between rhetoric as the amoral art of discovering any means of persuasion in a given context and rhetoric as a product to which ethical standards apply (Rowland & Womack, 1985). Indeed, he wrote that "we can prove people to be friends or enemies; if they are not, we can make them out to be so" (Aristotle, 1984, p. 2202), but viewed rhetors who utilize dishonest strategies as morally bad people (Rowland & Womack, 1985). In other words, just because a persuasive tactic would be effective does not mean it is morally correct to deploy it. In this study, we found evidence that beyond moral concerns, deploying problematic identification strategies—in this case, character attacks—can also prove problematic to the persuasive efficacy of one's rhetoric. Although attack-as-identification strategies may be effective to a target audience, we have shown that they can engender a backlash that counters insular gains. To return to the campus speaker context, it seems that fears of negative reactions to a controversial speech are more warranted if a speaker uses character attacks than if they use identification strategies, and the platform they are given may be protected by free speech, but it will not protect them from consequences, which may include the subsequent desire to restrict their speech. Thus, choosing to bully from one's pulpit may induce their audience to identify with them, but at a potential cost to both the speaker and to positive democratic outcomes.


Abramowitz, A. I., & Webster, S. W. (2018). Negative partisanship: Why Americans dislike parties but behave like rabid partisans. Advances in Political Psychology, 39(1), 119-135.

Aristotle. (1984). Rhetoric. In J. Barnes (Ed.), The complete works of Aristotle: The revised Oxford translation (pp. 2152-2269). Princeton, NJ: Princeton University Press.

Aspriadis, N., Takas, E., & Samaras, A. N. (2020). Country reputation assassination during the Greek memorandum re-negotiations. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 236-250). London/NY: Routledge.

Barnidge, M. (2017). Exposure to political disagreement in social media versus face-to-face and anonymous online settings. Political Communication, 34, 302-321.

Bauer-Wolf, J. (2019, September 16). Free speech laws mushroom in wake of campus protests. Inside Higher Ed.

Beck, A. T. (1999). Prisoners of hate: The cognitive basis of anger, hostility, and violence. Harper-Collins Publishers.

Benoit, W. A. (2000). Comparing the Clinton and Dole advertising campaigns: Identification and division in 1996 presidential television spots. Communication Research Reports, 17(1), 39-48.

Bimber, B. (2014). Digital media in the Obama campaigns of 2008 and 2012: Adaptation to the personalized political communication environment. Journal of Information Technology & Politics, 11(2), 130-150.

Black, E. (1970, April). The Second Persona. The Quarterly Journal of Speech, 56(2), 109-119.

Branaman, A. (1994). Reconsidering Kenneth Burke: His contributions to the identity controversy. The Sociological Quarterly, 35(3), 443-455.

Burke, K. (1937). Attitudes toward history. Berkeley, CA: University of California Press.

Burke, K. (1969). A rhetoric of motives (California ed.). Berkeley and Los Angeles, CA: University of California Press.

Burke, K. (1972). Dramatism and development. Worcester, MA: Clark University Press.

Chadwick, A., & Stromer-Galley, J. (2016) Digital media, power, and democracy in parties and election campaigns: Party decline or party renewal? The International Journal of Press/Politics, 21(3), 283-293.

Charland, M. (1987). Constitutive rhetoric: The case of the peuple québécois. Quarterly Journal of Speech, 73(2), 133-150.

Cheney, G. (1983). The rhetoric of identification and the study of organizational communication. Quarterly Journal of Speech, 69(2), 143-158.

Connaughton, S., & Jarvis, S. (2004). Apolitical politics: GOP efforts to foster identification from Latinos 1984-2000. Communication Studies, 55(3), 464-480.

Coombs, T., & Holladay, S. (2020). Corporate character assassination and crisis communication. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 225-235). Routledge.

Cram, T. (2017). "An open door": Responsibility and the comic frame in Obama's foreign policy rhetoric on Iran. Rhetoric & Public Affairs, 20(1), 69-108.

Daily Wire. (2019, February 1). Largest Christian university in U.S. blocks Shapiro from speaking. The Daily Wire.

Dean, A., Voss, D., & Draguljic, D. (2017). Design and analysis of experiments (2nd ed.). Berlin, Germany: Springer.

DellaPosta, D., Shi, Y., & Macy, M. (2015). Why do liberals drink lattes? American Journal of Sociology, 120(5), 1473-1511.

Egorova, E. & Egorova, E. (2020). Character assassination in the Soviet Union and Russia. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 92-104). London/New York: Routledge.

Ekins, E. (2017, October 31). The state of free speech and tolerance in America. CATO Institute.

Elsbach, K., & Bhattacharya, C. B. (2001). Defining who you are by what you are not: Organizational disidentification and the National Rifle Association. Organization Science, 12(4), 393-413.

Everson, D. H. (1982). The decline of political parties. Proceedings of the Academy of Political Science, 34(4), 49-60.

Foss, S. K., & Griffin, C. L. (1995) Beyond persuasion: A proposal for an invitational rhetoric. Communication Monographs, 62, 2-18.

Furnham, A., Richards, S., Rangel, L., & Jones, D. N. (2014). Measuring malevolence: Quantitative issues surrounding the Dark Triad of personality. Personality and Individual Differences, 67, 114-121.

Gendron, S. (2012). Exploiting the Hutu/Tutsi divide: The relationship between extremist propaganda and genocide in Rwanda. In M. Vuorinen (Ed.), Enemy images in war propaganda (pp. 89-106). Newcastle upon Tyne, UK: Cambridge Scholars Publishing.

Gilbert, C. J. (2020). Argumentum ad carricare as a mode of character attack: Alec Baldwin as Donald Trump on Saturday Night Live. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 209-224). London/New York: Routledge.

Greene, S. (1999). Understanding party identification: A social identity approach. Political Psychology, 20(2), 393-403.

Greene, S. (2004). Social identity theory and party identification. Social Science Quarterly, 85(1), 136-153.

Haberman, M., & Shear, M. D. (2019, March 21). Trump signs executive order protecting free speech on college campuses. The New York Times.

Halperin, E. (2008). Group-based hatred in intractable conflict in Israel. Journal of
Conflict Resolution
, 52(5), 713-736.

Hayes, A. F. (2018). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (2nd ed.). New York, NY: Guildford Press.

Heclo, H. (2000). Campaigning and governing: A conspectus. In N. J. Ornstein & T. E. Mann (Eds.), The permanent campaign and its future (pp. 1-37). Washington, DC: American Enterprise Institute and Brookings Institution.

Hibbing, J. R., Smith, K. B., & Alford, J. R. (2013). Predisposed: Liberals, conservatives, and the biology of political differences. London/New York: Routledge.

Icks, M., & Shiraev, E. (2014). Introduction. In M. Icks & E. Shiraev (Eds.), Character assassination throughout the ages (pp. 1-13). New York, NY: Palgrave Macmillan.

Icks, M., Shiraev, E., Keohane, J., & Samoilenko, S. A. (2020). Character assassination: Theoretical framework. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 11-24). London/New York: Routledge.

Iyengar, S., & Westwood, S. J. (2015). Fear and loathing across party lines: New evidence on group polarization. American Journal of Political Science, 59(3), 690-707.

Jennings, F. J., Coker, C. R., Bramlett, J. C., Reed, J. L., & Bolton, J. P. (2018). Late night with Donald Trump: An exploration of the combined effects of political comedy and political advertising. In B. R. Warner, D. G. Bystrom, M. S. McKinney, & M. C. Banwart (Eds.), An unprecedented election: Media, communication, and the electorate in the 2016 campaign (pp. 235-252). Santa Barbara, CA: ABC-CLIO, LLC.

Johnston, C. D., Lavine, H. G., & Federico, C. M. (2017). Open versus closed: Personality, identity, and the politics of redistribution. Cambridge, UK: Cambridge University Press.

Jones, J. M. (2018, March 12). More U.S. college students say campus climate deters speech. Gallup.

Keen, S. (1991). Faces of the enemy: Reflections of the hostile imagination. New York, NY: Harper & Row.

Kim, Y. M., Hsu, J., Neiman, D., Kou, C., Bankston, L., Kim, S. Y., & Raskutti, G. (2018). The stealth media? Groups and targets behind divisive issue campaigns on Facebook. Political Communication, 35(4), 515-541.

Knight Foundation. (2018, March 12). Free expression on campus: What college students think about first amendment issues.

Lemert, C. (2018). Introduction. In C. Lemert (Ed.), Social theory: The multicultural, global, and classic readings (6th ed.) (pp. 1-18). London/New York: Routledge.

Levendusky, M. (2009). The partisan sort: How liberals became Democrats and conservatives became Republicans. Chicago, IL: The University of Chicago Press.

Levendusky, M. S. (2013). Why do partisan media polarize viewers? American Journal of Political Science, 57(3), 611-623.

Levendusky, M., & Malhotra, N. (2016). Does media coverage of partisan polarization affect political attitudes? Political Communication, 33, 283-301.

Levitsky, S., & Ziblatt, D. (2018). How democracies die. London, UK: Penguin Books.

Lupia, A. (2013). Communicating science in politicized environments. Proceedings of the National Academy of Sciences.

Martin, J. (2008, October 10). McCain booed defending Obama to angry crowd. Politico.

McLeod, D. M. & Detenber, B. H. (1999). Framing effects of television news coverage of social protest. Journal of Communication, 49(3), 3-23.

Min Kim, S., & Gearan, A. (2018, August 30). Jared Kushner ramps up push for criminal justice reform. The Washington Post.

Minielli, M. (2020). Character assassination and the Nixon White House. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 282-294). London/New York: Routledge.

Muddiman, A., & Stroud, N. (2017). News values, cognitive biases, and partisan incivility in comment sections. Journal of Communication, 67(4), 586-609.

Munro, G. D., Weigh, C., & Tsai, J. (2010). Motivated suspicion: Asymmetrical attributions of the behavior of political ingroup and outgroup members. Basic and Applied Social Psychology, 32, 173-184.

Orwell, G. (1968). Politics and the English language. In S. Orwell & I. Angos (Eds.), The collected essays, journalism and letters of George Orwell (pp. 127-140). San Diego, CA: Harcourt, Brace, Javanovich.

Overington, M. A. (1977). Kenneth Burke as social theorist. Sociological Inquiry, 47(2), 133-141.

Paolacci, G., Chandler, J., & Ipeirotis, P. G. (2010). Running experiments on Amazon Mechanical Turk. Judgment and Decision Making, 5(5), 411-419.

Pariser, E. (2011). The filter bubble: What the internet is hiding from you. London, UK: The Penguin Press.

Pettigrew, T. F. (1979). The ultimate attribution error: Extending Allport's cognitive analysis of prejudice. Personality and Social Psychology Bulletin, 5, 461-476. doi:10.1177/ 014616727900500407

Pew Research Center. (2017, October 5). The partisan divide on political values grows even wider.

Pew Research Center. (2019, February 4). Public’s policy priorities for 2019.

Riccardi, N., & Fingerhut, H. (2019, January 4). AP-NORC poll: Immigration among the top concerns in 2019. The Associated Press.

Roderick, M. (2014). From identification to dis-identification: Case studies of job loss in professional football. Qualitative Research in Sport, Exercise, and Health, 6(2), 143-160.

Rodrigo, C. M. (2018, December 21). Trump signs criminal justice overhaul. The Hill.

Rowland, R. C., & Womack, D. F. (1985). Aristotle's view of ethical rhetoric. Rhetoric Society Quarterly, 15(1/2), 13-31.

Schneider, E. (2019, November 12). Buttigieg woos ‘future former Republicans’ in wide-open N.H. primary. Politico.

Shiraev, E. (2014). Character assassination: How political psychologists can assist historians. In M. Icks & E. Shiraev (Eds.), Character assassination throughout the ages (pp. 15-33). New York, NY: Palgrave Macmillan.

Shiraev, E., & Makhovskaya, O. (2020). The traumatic psychological impact of character attacks on targets. In S. A. Samoilenko, M. Icks, J. Keohane, & E. Shiraev (Eds.), Routledge handbook of character assassination and reputation management (pp. 36-44). Routledge.

Smart, J., & Shiraev, E. (2014). Character attacks and American presidents. In M. Icks & E. Shiraev (Eds.), Character assassination throughout the ages (pp. 215-235). New York, NY: Palgrave Macmillan.

Sullum, J. (2018, October 17). Kanye’s real success: Trump now backs criminal justice reform. New York Post.

Trump, D. (2020, February 28). Charleston, South Carolina rally transcript. Rev.

Valentine, S. R., Hanson, S. K., & Fleischman, G. M. (2019). The presence of ethics codes and employees' internal locus of control, social aversion/malevolence, and ethical judgment of incivility: A study of smaller organizations. Journal of Business Ethics, 160(3), 657-674.

Verkuyten, M. & Yildiz, A. A. (2007). National (dis)identification and ethnic and religious identity: A study among Turkish-Dutch Muslims. Personality and Social Psychology Bulletin, 33, 1448-1462.

Wagner, M. W., Mitchell, D., & Theiss-Morse, E. (2011). The consequences of political vilification [2011 Annual meeting paper]. American Political Science Association.

Wander, P. (1984). The third persona: An ideological turn in rhetorical theory. Central States Speech Journal, 35(4), 197-216.

Warner, B. R., Galarz, R., Coker, C. R., Tschirhart, P., Hoeun, S., Jennings, F. J. & McKinney, M. S. (2019). Comic agonism in the 2016 campaign: A study of Iowa Caucus rallies. American Behavioral Scientist, 63(7), 836-855.

Warner, B. R., & Villamil, A. (2017). A test of imagined contact as a means to improve cross-partisan feelings and reduce attribution of malevolence and acceptance of political violence. Communication Monographs, 84(4), 447-465.

Weber Shandwick. (2019, June 26). Civility in America 2019: Solutions for tomorrow.

Wike, R. (2016, October 12). Americans more tolerant of offensive speech than others in the world. Pew Research Center.

Wike, R., & Simmons, K. (2015, November). Global support for principle of free expression, but opposition to some forms of speech. Pew Research Center.

Yang, J., Rojas, H., Wojcieszak, M., Aalberg, T., Coen, S., Curran, J., Hayashi, K., Iyengar, S., Jones, P. K., Mazzoleni, G., Papathanassopoulos, S., Rhee, J. W., Rowe, D., Soroka, S., & Tiffen, R. (2016). Why are "others" so polarized? Perceived political polarization and media use in 10 countries. Journal of Computer-Mediated Communication, 21, 349-367.


* Email:

ISSN: ISSN 2398-5836

Copyright (c) 2021 Amy Schumacher-Rutherfored, Ashley Muddiman