Perceptions of Uncivil Discourse Online: An Examination of Types and Predictors

Incivility in Public Discourse: Content, Causes, and Effects

The concepts of civility and incivility, Herbst (2010) has rightly noted, are “very much in the eye of the beholder” (p. 3). In other words, what is perfectly civil to one person might strike another person as perfectly uncivil. Nevertheless, the range of messages typically defined as uncivil by scholars have some common themes. For instance, ad hominem attacks are often cited, as are vulgarity and various forms of exaggeration (e.g., Berry & Sobieraj, 2014; Borah, 2013; Jamieson, 1997; Muddiman, 2013; Santana, 2014). As Massaro and Stryker (2012) concluded in their extensive review of the literature on incivility, “identifying civil and uncivil political discourse is not impossible, even though we probably will never achieve complete consensus about precisely where to draw the line” (p. 440). In the present study, we adopt Coe et al.’s (2014) definition of incivility as “features of discussion that convey an unnecessarily disrespectful tone toward the discussion forum, its participants, or its topics” (p. 660). For our purposes, this definition has two key advantages. First, it is consistent with the nature of online forums. Indeed, the Coe et al. study focused directly on newspaper website comments, ensuring that the conceptual definition encompasses that environment and that the operational definition includes components of incivility likely to be present in such forums. Second, the key features of the definition are consistent with those employed in prior research.

Defined along these lines, it is abundantly clear that incivility is—and has long been—quite common in American public discourse (see Shea & Fiorina, 2013). In the domain of mediated political speech, for example, Sobieraj and Berry (2011) content analyzed a wide range of cable television and talk radio programs, political blogs, and newspaper columns and found that nearly 90% of their cases included some form of “outrage” (a construct related to incivility, but not synonymous with it). In online discussions in particular, incivility can be quite common as well, though there is variation in the literature on this point. For example, Coe et al. (2014) analyzed comments posted to the Arizona Daily Star and found that roughly one in five comments included some form of incivility, with direct insults of another person far and away the most common variety. In contrast, Ruiz et al. (2011) examined the New York Times (among other international sources) and found its discussions were largely devoid of direct insults, though broader kinds of “derogatory language” occurred in roughly 3% of the comments. Notably, the presence of incivility may sometimes come at the expense of explicitly civil forms of communication. For example, da Silva (2013) found several characteristics of civil discourse, such as cooperation and explicit displays of respect, occurred infrequently in comments posted to political stories on two Portuguese newspaper websites.

The literature has also considered some of the causes of incivility, focusing in particular on structural factors. For example, in their extensive study of “the new incivility,” Berry and Sobieraj (2014) point to a variety of economic, regulatory, and technological changes that have produced a media system “uniquely supportive of outrage-based political opinion” (p. 75). In particular, the rise of partisan niche media has created a forum that emphasizes and rewards forms of incivility such as ad hominem attacks (Jamieson & Hardy, 2012; Massaro & Stryker, 2012). In the context of online environments specifically, smaller scale features often influence levels of incivility. For example, hard news stories tend to generate higher levels of incivility in the accompanying comments, as do articles that prominently feature familiar politicians who have an identifiable affiliation (Coe et al., 2014). Anonymity matters as well; Santana (2014) found incivility in approximately 53% of comments posted anonymously to news websites, whereas only roughly 29% of the nonanonymous comments contained incivility. Notably, spending time in online discussion environments where uncivil behavior is common can encourage participants to believe such behavior is acceptable, which in turn increases the likelihood that they themselves will engage in that behavior (Hmielowski, Hutchens, & Cicchirillo, 2014). Some forms of active moderation (e.g., reporter participation in the discussion forum), however, can decrease incivility (Stroud, Scacco, Muddiman, & Curry, 2015).

Whatever its cause, incivility has effects on those who encounter it—some of which are negative. For example, incivility can harm perceptions of political arguments (e.g., Brooks & Geer, 2007; Mutz, 2007), weaken political trust (Mutz & Reeves, 2005), hurt the favorability ratings of political institutions (Muddiman, 2013), and heighten perceived political polarization among the public (Hwang, Kim, & Huh, 2014). More broadly, Maisel (2012) explains how incivility might have deleterious effects on the political process by discouraging talented candidates from seeking (re)election, as well as by making compromise—and, consequently, governance—more difficult. In computer-mediated contexts, specifically, incivility in blog comments can influence people’s response to the topic under discussion (Anderson et al., 2014). Borah (2013, 2014), for instance, conducted studies of students to see how uncivil blog commentary interacted with news frames. These studies found incivility to decrease open-mindedness, as well as political trust and efficacy. At the same time, however, incivility increased online participation and, in some cases, the perceived credibility of accompanying news stories (see also Gervais, 2015; Thorson, Vraga, & Ekdale, 2010). In sum, there is little question that incivility matters—sometimes in a positive fashion, but often in more concerning ways.

Public Perceptions of Uncivil Messages

What virtually all of the literature reviewed above has in common is that it attends primarily to elite and/or political discourse as the site or target of incivility. The relatively few studies that somewhat depart from this norm (e.g., Anderson et al., 2014; Borah, 2013, 2014) tend to focus on the consequences of incivility for those who encounter it rather than on how incivility itself is perceived. The focus on politics and elites in prior research is understandable, but leaves a considerable gap in the literature so far as everyday moments of incivility (i.e., those created and encountered by regular citizens) are concerned. We seek to address this void by exploring how the public perceives incivility, looking in particular at reactions to different types of content that have been defined by scholars as uncivil and the individual-level characteristics that influence perceptions of incivility.

Our first interest in gaining insight into public perceptions of incivility is in whether people draw distinctions between different types of incivility. Most studies of incivility operationalize the concept as specific features or behaviors that can be systematically tracked. Yet, these various features might elicit different reactions among the public. To explore this possibility, we again rely on the Coe et al. (2014) definition because their operational definition drew heavily from several other studies of incivility (e.g., Jamieson, 1997; Sobieraj & Berry, 2011). Coe et al. tracked five specific forms of incivility: name-calling (i.e., ad hominem attacks), vulgarity (i.e., curse words, including minor ones such as “damn”), lying accusations (i.e., charges that another person is dishonest), pejoratives for speech (i.e., mocking the way someone expresses himself or herself), and aspersion (i.e., derisive/disrespectful attacks on an idea). If people perceive these forms of content as uncivil, they should view them differently than messages that do not contain such content. Accordingly, we propose a hypothesis:

We reasoned further that, if people are likely to perceive certain messages as more uncivil than others, this would likely happen among different types of uncivil content as well. That is, there are probably certain types of incivility that are more obviously uncivil than others. We expect that one such type will be name-calling. Coe et al. (2014) found name-calling to be the most common form of incivility in comments to online news articles, occurring 14% of the time. No other form of incivility occurred in more than 3% of comments. Name-calling is also a feature cited in nearly every extant study of incivility (e.g., Berry & Sobieraj, 2014; Jamieson, 1997; Massaro & Stryker, 2012). The prominence of name-calling in public discussion and scholarly definitions leads us to believe that it is among the more familiar and obvious forms of incivility. If so, we expect that people will readily identify it when they see it, and label it as uncivil. We therefore propose two additional hypotheses:

Our final interest is in the individual-level characteristics that might influence perceptions of incivility. We explore three general categories of individual-level characteristics that could be systematically associated with incivility perceptions—demographics, personality traits, and news habits—providing what is to our knowledge the broadest analysis of such predictors to date.

First, demographics provide some fundamental variation that might matter for perceptions of incivility, as they do for a host of other perceptions, attitudes, and behaviors. In online environments, for instance, there is some evidence that men post more negative comments than do women (Moss-Racusin, Molenda, & Cramer, 2015). To the standard demographic battery (e.g., sex, age, race/ethnicity, etc.), we add geographic region of the United States (e.g., South). Given that interpersonal communication norms vary regionally (Andersen, Lustig, & Andersen, 1987), it seems possible that perceptions of incivility might as well. Furthermore, there is some evidence that workplace incivility is perceived as more harmful in the northern United States than in the south (Kelley, 2007). One of the two samples we use herein is drawn nationally and offers substantial geographic variation, so it is well suited to exploring the possibility that regional variation helps explain perceived incivility.

Second, we examine personality traits, drawing on two key measures. The first is the “Big Five,” a widely used metric for examining how personality influences a whole range of perceptions, attitudes, and behaviors (Gosling, Rentfrow, & Swann, 2003). Notable correlations between these personality characteristics and incivility might suggest that incivility is a chronic trait as much as it is a temporary state of being. In addition, certain components of the Big Five seem especially likely to connect to perceptions of incivility. For example, agreeableness predicts empathy and nonretaliation (Ashton, Paunonen, Helmes, & Jackson, 1998), which might make a feature of incivility such as name-calling seem particularly off-putting. Employing the Big Five gives us a broadly understood personality baseline, while our second measure—conflict avoidance—allows us to measure a personality characteristic that has been tied more directly to incivility per se. Indeed, inclination toward conflict avoidance has been included as a control variable in some past analyses of incivility (e.g., Ben-Porath, 2010; Muddiman, 2013), and one study found that it interacted with incivility to help explain levels of political trust (Mutz & Reeves, 2005).

Finally, we consider news consumption habits. Given that news media are a common venue in which people might encounter incivility (Sobieraj & Berry, 2011), it seems possible that different degrees or kinds of news exposure might contribute to how people perceive incivility. York (2013), for instance, found that viewing cable television news was a significant predictor of characterizing “the overall tone and level of civility in politics today” as “mostly negative” (p. 117). Gervais (2014), meanwhile, found that use of uncivil political media can lead people to express themselves in less civil ways. Gervais’s finding may suggest a process of desensitization is at work. Scholars interested in media violence have long known that repeated exposure to mediated acts of violence desensitizes individuals, emotionally and physiologically, to such violence (for a review, see Harris, 2009). Incivility might function similarly. Enough exposure and a certain message may no longer seem so uncivil, which could in turn lead people to be more likely to espouse similar messages themselves.

Taken together, an analysis of demographics, personality traits, and news habits should provide considerable insight into the characteristics that help explain perceived incivility. Given the lack of consistent, directional findings about how these various characteristics might relate to perceptions of routine incivility, we pose the following research question:

Method

Procedure and Materials

Participants were first asked to report demographic information and complete measures of personality traits, conflict avoidance, and media exposure. Participants were then informed that they would see written comments that were posted by readers to articles appearing on a newspaper’s website. Participants read a series of 12 statements and, after each statement, completed four items (described in the following section) used to evaluate their perceptions of the relative civility of each statement. To minimize the possibility that the particularities of any single message were responsible for the observed effects, two statements were used to represent each of the five types of incivility. An additional two statements were used as control statements in which uncivil speech acts were not present. All 12 statements, which are presented in Table 1, were taken directly from Coe et al.’s (2014) content analysis of incivility in discussions of online news. Given that the present study was devised as a survey rather than as an experiment, all participants evaluated all 12 statements. The order of the 12 statements was randomized for each participant to minimize the possibility of an order effect.

OPEN IN VIEWER

Table 1.

Items Used to Evaluate the Different Classes of Incivility.

Type of incivility	Statements
Name-calling	Just because something is not popular with the beer swilling Texas drawling Nascar crowd here does not mean that the rest of the world agrees that it is worthless.
	At least the morons in the state capital no longer have control of this process!
Vulgarity	I hope the voters will kick that politician out on his pompous ass next election.
	That person said not one damn word in that article. All BS just like the politicians—the same crap.
Lying accusation	Americans have been screaming at the top of their lungs that this government is wrong, is corrupt, is lying, is deceiving the people, and is violating our constitution.
	We need to get everyone out of office and start fresh. Make it so that lawyers cannot run for office in the executive or legislative branches of government. They lie and should not be trusted.
Pejorative for speech	Quit crying over spilled milk.
	I am sick and tired of people throwing tantrums.
Aspersion	Our justice system is just as corrupt and lousy as any in the world.
	Texting while driving is stupid.
No incivility (control statements)	95% of our students qualify for free or reduced lunches???? What in the world does that have to do with being able to pass a test. [sic] “Poor” is no excuse.
	The article was not correct with regard to the personnel required to fly and maintain the unmanned aerial vehicles. The article stated that three individuals were required to fly each UAV. The real number is two (Pilot and Sensor Operator).

Note. All incivility statements were taken from Coe, Kenski, and Rains (2014).

Measures

Incivility was measured using four semantic differential items created by the authors. The four items were rated on 7-point scales with the following anchors: civil/uncivil, rude/polite (reverse coded), necessary/unnecessary, and disrespectful/respectful (reverse coded). These items were randomly rotated and the valences of the anchors were mixed to avoid negativity or positivity response bias. The mean of the four items was computed to evaluate the relative perceived incivility; larger scores indicate greater perceptions of incivility. The mean level of incivility across all 12 of the statements was 4.23 (SD = .84). Cronbach’s alpha was computed to examine the internal consistency of the perceived incivility measure. As shown in Table 2, the alphas were above .80 for all five types of incivility. We opted to use this four-part approach to measuring incivility, as opposed to a single item, for three reasons. First, the items connect directly to the various elements of our conceptual definition of incivility. Although not all scholars agree on the definition of incivility, our conceptual definition includes most of the dimensions emphasized in the literature, and our operationalization flows logically from those dimensions. Second, there is potential “folly of single indicants” (Jacoby, 1978, p. 93) when single items are used. Measurement tools are more psychometrically sound when they utilize related but different items to capture a concept because the error stemming from any single item tends to be mitigated by the signal projected from the other items. Third, at an empirical level, the fact that the alphas were above .80 for all five types of incivility and the two control statements suggests that the public perceives these items in a consistent and coherent way.

OPEN IN VIEWER

Table 2.

Mean Ratings for the Five Different Types of Incivility.

Type of incivility	α	M	SD
Name-calling	.88	5.05 a	1.24
Vulgarity	.90	5.00a,d	1.33
Lying accusation	.86	3.99 c	1.30
Pejorative for speech	.86	3.92b,c	1.16
Aspersion	.80	3.74 b	1.23
Control Statement 1	.93	4.77 d	1.76
Control Statement 2	.93	2.69	1.42

Note. All groups are significantly different from one another unless labeled with similar superscripts.

Demographic variables

Several demographic variables were measured and evaluated as predictors of incivility. Descriptive information for respondents’ sex, race, education (in years), and income (in thousands) was provided in describing the sample. A dichotomous measure of conservative political ideology was constructed by assigning respondents who self-reported being very conservative and somewhat conservative a score of 1 and participants who did not indicate being very or somewhat conservative a score of 0. A dichotomous measure of liberal political ideology was constructed by assigning respondents who reported being very or somewhat liberal a score of 1 and those who did not a score of 0. Finally, a series of dummy-coded variables were created to evaluate the geographic region of the United States in which respondents resided based on one of four categories: Northeast (n = 69), Midwest (n = 100), South (n = 136), and West (n = 93). The West was used as the reference group for the dummy-coded variables.

Personality characteristics

Respondents’ personality traits were evaluated with the 10-item brief measure of the Big Five personality domain inventory (Gosling et al., 2003). Respondents rated the degree to which the following traits apply to them using a 7-point scale with the anchors strongly disagree (1) and strongly agree (7): extraverted/enthusiastic, critical/quarrelsome, dependable/self-disciplined, anxious/easily upset, open to new experiences, reserved/quiet, sympathetic/warm, disorganized/careless, calm/emotionally stable, and conventional/uncreative. Means were computed to form two-item measures of extraversion (M = 4.08, SD = 1.49, r = .47), agreeableness (M = 5.19, SD = 1.17, r = .24), conscientiousness (M = 5.89, SD = 1.06, r = .30), emotional stability (M = 4.97, SD = 1.27, r = .45), and openness to new experiences (M = 5.04, SD = 1.10, r = .25).

Conflict avoidance was evaluated using a five-item measure used by Mutz and Reeves (2005) and Muddiman (2013) based on Goldstein’s (1999) work. Participants rated their agreement with the following statements on a 5-point scale with the anchors strongly disagree (1) and strongly agree (5): “I hate arguments,” “I find conflicts exciting” (reverse coded), “I enjoy challenging the opinions of others” (reverse coded), “Arguments don’t upset me” (reverse coded), and “I feel upset after an argument.” The mean of the items was computed to form the measure of conflict avoidance (M = 3.55, SD = 0.79, α = .81)

News consumption habits

Respondents’ news habits were evaluated with four items. Respondents reported the number of days during the past week in which they watched television news programs (M = 4.96, SD = 2.47), read a paper copy of a newspaper (M = 2.66, SD = 2.77), read an online newspaper (M = 2.01, SD = 2.65), and listened to talk radio shows that invite listeners to call in (M = 1.66, SD = 2.34).

Results

Predictors of Perceived Incivility

Research Question 1 asked about individual characteristics—demographics, personality traits, and news habits—that might be associated with perceived incivility. Hierarchical ordinary least squares (OLS) regression models were tested to answer this question. An overall model was tested in which the measures of perceived incivility for each of the five types of uncivil speech acts were combined into a single aggregate variable. A series of five regression models was then tested to evaluate each of the five types of incivility separately. The only difference in the models was the outcome variable (i.e., perceived incivility overall and perceived incivility for one of the five forms of uncivil speech acts). The variables were entered in three separate blocks. The demographic factors were entered in the first block, followed by the five personality characteristics and conflict avoidance in the second block, and the four measures of news consumption habits in the final block.

The results of the regression model for the aggregate measure of perceived incivility are reported in Table 3. Table 4 contains the results of the individual models for five types of uncivil speech acts. In the model examining the aggregate incivility measure, the block containing demographic variables predicted statistically significant but modest amounts of variance in incivility perceptions. In regard to individual predictors, females and those who scored higher on the personality trait of agreeableness were more likely to perceive the speech acts to be uncivil, whereas respondents who self-identified as being conservative were less likely to perceive the statements as being uncivil.

OPEN IN VIEWER

Table 3.

Results of the Regression Model Predicting Aggregate Perceptions of Incivility.

	Aggregate measure of perceived incivility
	B	ΔR2
Block 1: Demographics		.07**
Sex (female = 1, male = 0)	.21***
Race (White = 1, else = 0)	−.06
Education	.04
Income	.05
Conservative	−.12*
Liberal	.00
Region–Northeast	−.05
Region–Midwest	.00
Region–South	−.02
Block 2: Personality		.03 †
Extraversion	.05
Agreeableness	.17**
Conscientiousness	−.05
Emotional stability	.00
Openness	−.04
Conflict avoidance	.04
Block 3: News habits (days per week)		.03 †
TV news	−.06
Newspaper (print)	−.11 †
Newspaper (online)	.09
Talk radio	−.04

Note. The results for each block are reported when that block was added to the model. All “Region” variables are dummy-coded; the West served as the reference group. Model summary: F(19, 312) = 2.45, p < .001, R² = .13.

†

p < .10. *p < .05. **p < .01. ***p < .001.

OPEN IN VIEWER

Table 4.

Results of the Regression Models Predicting Perceptions of the Five Types of Uncivil Speech.

	Form of incivility
	Name-calling	Vulgarity	Lying accusation	Pejorative for speech	Aspersion
	β	β	β	β	β
Block 1: Demographics
Sex (female = 1, male = 0)	.22***	.21***	.13**	.15**	.04
Race (White = 1, else = 0)	.10 †	−.03	−.08	−.08	−.10 †
Education	.07	.04	.05	.00	−.03
Income	.01	.06	.10 †	−.01	.03
Conservative	−.08	−.10	−.16**	−.07	−.04
Liberal	.00	.00	.08	−.04	.00
Region–Northeast	−.06	−.12 †	−.05	.04	.02
Region–Midwest	−.04	−.01	.01	.06	−.01
Region–South	−.07	−.04	−.02	.05	.04
Block 2: Personality traits
Extraversion	.10	.07	.03	.06	−.03
Agreeableness	.16*	.24***	.06	.07	.09
Conscientiousness	−.08	−.02	−.05	−.03	.01
Emotional stability	−.04	.03	.05	.00	−.01
Openness	.07	−.02	−.03	−.07	−.07
Conflict avoidance	.05	.06	−.02	.08	−.02
Block 3: News habits (days per week)
TV news	−.01	−.04	−.06	−.08	−.09
Newspaper (print)	−.13*	−.06	−.11 †	.01	−.07
Newspaper (online)	.10 †	.08	.05	.02	.05
Talk radio	−.07	−.02	.02	.03	−.09

Note. The results for each block are reported when that block was added to the model. All “Region” variables are dummy-coded; the West served as the reference group. Model summaries: Name-calling, F(19, 312) = 3.01, p < .001, R² = .16; Vulgarity, F(19, 311) = 3.03, p < .001, R² = .16; Lying accusation, F(19, 311) = 2.09, p < .01, R² = .11; Pejorative for speech, F(19, 311) = 1.07, p = .38, R² = .06; Aspersion F(19, 312) = 0.89, p = .59, R² = .05.

†

p < .10. *p < .05. **p < .01. ***p < .001.

In terms of individual forms of uncivil speech acts, respondents’ sex was the only consistent predictor of incivility perceptions. None of the predictors reached conventional levels of significance for perceptions of incivility in the messages containing aspersion. The aggregate finding that those who rated highly on agreeableness were more likely to evaluate the statements as uncivil was drawn from the name-calling and vulgarity statements. The aggregate finding that conservatives were less likely to see the statements as uncivil was affected primarily by the lying accusation category. Conservatives did not see statements referring to others as liars as uncivil as did liberals and moderates. Print newspaper consumption was associated with decreases in incivility perceptions for name-calling.

Discussion

This study sought to contribute to the growing literature on incivility by exploring public perceptions of the various types of incivility that people might be likely to encounter in online environments. Our results indicate that, just as arguments can be considered strong or weak based on structural and grammatical features (Areni, 2003), different classes of uncivil discourse are evaluated differently when it comes to civility perceptions. We examined five forms of incivility identified in past research and found that participants saw considerable variation in the extent of incivility present in those different speech acts. We predicted that there would be differences in perceptions of the five types of uncivil discourse, and this hypothesis was supported. We further hypothesized that name-calling would be considered more uncivil than would other forms of incivility, which was mostly supported in the data. Among middle-aged adults, name-calling received the highest incivility rating of the five forms of incivility examined; it was, however, statistically indistinguishable from vulgarity, which was also rated as highly uncivil and which was viewed as the single most uncivil form among college students. Our assumption that name-calling would be considered the most uncivil type of incivility was driven by the recognition of its frequency. In contrast, vulgarity might be an easily identifiable form of incivility because of its infrequency. That is, some forms of incivility might be infrequent because people share the norm that these types of utterances are usually unacceptable forms of speech in public arenas. When such norms are broken, people notice. Vulgarity, of course, does sometimes appear in online discussions, probably because the cloak of anonymity that online forums provide can release some individuals from feelings of responsibility or consequence for their verbal expressions (O’Sullivan & Flanagin, 2003).

Our findings also provide significant insight into the individual characteristics that influence perceptions of incivility. Of the demographic, personality, and news consumption variables examined, only being female was consistently associated with rating statements as uncivil, though other traits predicted perceptions of certain types of incivility. Specifically, being female was associated with the ratings of all forms of incivility except aspersion. The fact that females stood out so noticeably across different models of incivility perceptions here warrants consideration. Clearly, there are different social norms for men and women when it comes to expression, which may explain the greater sensitivity to incivility by females in certain contexts. Differences in expression appear to start early, with boys expecting less disapproval from their parents for aggression than do girls; boys also expect less guilt and upset for aggression than do girls (Perry, Perry, & Weiss, 1989). As noted above, recent research found that women were less likely to offer negative comments in an online forum (Moss-Racusin et al., 2015). Our findings suggest that this may be explained, in part, by perceived incivility. If women have a heightened sensitivity to uncivil comments, they may be less willing themselves to engage in such discourse. Future research could test this relationship more directly.

Although not as consistent as the association between sex and incivility evaluations, the personality trait agreeableness was also associated with incivility evaluations overall in the national sample. Those whose personalities were high on the agreeableness trait were more likely to see comments as uncivil. It may be the case that those who have agreeable personalities are more sensitive to potential disruptions to social harmony and are thus more likely to monitor comments for their potential impacts on the conversation. This idea is broadly consistent with the idea that agreeableness predicts empathy and nonretaliation (Ashton et al., 1998).

Print newspaper consumption was associated negatively with the incivility ratings for the name-calling statements. In other words, print news consumers appear to have a higher threshold for incivility than do other types of news consumers and nonconsumers. This would at first glance seem to be consistent with the idea that audiences who encounter incivility become desensitized to it, except that only one form of media use predicted this relationship—and it was a form that some have found to be less likely to contain incivility (Sobieraj & Berry, 2011). Given that, another possible explanation is that it may be more difficult to avoid counter-attitudinal positions when reading print stories than ones online. Consequently, people engaging in print consumption may be more used to exposure to ideological content from the other side and thus not see strident remarks as problematic. If print newspaper reading is associated with higher levels of tolerance, future research should unpack whether such consumption causes greater tolerance or whether the politically tolerant are more attracted to forums that provide more opportunities for counter-attitudinal exposure.

Taken as a whole, the results related to the research question demonstrate that an audience’s perceptions of uncivility are not likely to be uniform. Scholars and practitioners cannot assume that all audience members will interpret an uncivil act in the same way. Indeed, a range of background characteristics—sex, ideology, the personality trait of agreeableness—may influence a given audience member’s response to incivility. These differences may not be limited to perceptions and could extend to behavioral responses to incivility. Consequently, researchers who want to design interventions that promote citizen deliberation would be well-served to take into account various background characteristics in their designs.

The specific perceptions of the “uncivil” control statements we employed provide insights that should help support future research in this area. We purposely chose two very different types of control statements. One contained a strong, declarative position that we suspected would not be shared by a significant percentage of participants in the study. The second was declarative in its composition but focused on a statistical fact rather than a political position. Although the incivility category statements were perceived as more uncivil than the fact-focused control statement, the control statement that was declarative about a political position received a high incivility rating—one that was lower than name-calling but was statistically comparable to vulgarity (and higher than the other types). What these mixed findings from the different control statements seem to suggest is that although incivility should be considered distinct from disagreement conceptually, it may not be perceived by members of the public as such. Disagreement is an important part of healthy deliberation, which is needed for our government to make the best societal decisions possible. Yet, individuals have a hard time coping with information that is not consistent with the way they see the world—and this can shape their perceptions of incivility, especially along political lines (cf. Muddiman, 2013). Future research should focus on this paradox and identify ways that people can recognize differences of opinion without deciding those differences have no place in political discussion. Indeed, when one says that a comment is uncivil, he or she is making the claim that the comment is inappropriate. It is therefore imperative that citizens are able to make a distinction between incivility and disagreement because a representative democracy cannot function without tolerance for disagreement.

It should also be noted that the declarative position contained four question marks in a row that might have been interpreted as hyperbolic or extreme. Sobieraj and Berry (2011) contend that textual presentation such as “the deliberate use of uppercase letters, multiple exclamation points, enlarged text, and so on” (p. 40) is the digital equivalent of shouting (see also Gervais, 2015). Most people dislike being shouted at, so participants may have assessed this pattern as incivility. Future research should consider the extent to which grammatical format affects people’s interpretations of incivility and broader forms of information online.

This study was not without limitations. First, our samples were targeted to specific age groups, which limits the generalizability of them. The national sample from our primary survey was representative of 40- to 65-year-old Americans. Because the results were largely similar across both samples, we believe that they capture meaningful insight into how people perceive incivility.

Second, fixed message effects might be considered a study limitation. We should note, however, that we included multiple messages for each form of incivility and two control statements—a design which should help limit perceptions being influenced by specific characteristics of any single message. Nonetheless, we still assessed just 12 different statements, and the topics and structure of the statements necessarily differed by category. Future research could construct otherwise similar messages that included different types of incivility as a way to better control for the possibility of message effects.

Third, this study did not examine social media consumption as a potential media correlate. It was excluded in large part because social media involvement includes so much that falls outside of the realm of traditional news consumption, and scholarly work suggests users and researchers often disagree over how to categorize news content (Vraga, Bode, Smithson, & Troller-Renfree, 2016). Nevertheless, it is clear that as the media environment continues to evolve, future research will need to consider the different uses of social media and how those uses may affect interpretations and tolerance of incivility.

Finally, although aspersion has been identified as a form of incivility in past research (e.g., Coe et al., 2014; Jamieson, 1997), herein it received the lowest incivility evaluation of the five forms. Furthermore, no demographic characteristics, psychological traits, or media exposure behaviors were associated with aspersion evaluations. It may be the case that the statements used to test aspersion rating were not as provocative as is typical. In particular, one of the aspersion statements was “Texting while driving is stupid.” We did not ask survey participants how they felt about people who text while driving, but we suspect that most people would agree that it is a bad and dangerous activity. Their agreement with the statement may have mitigated their incivility perceptions, which is consistent with the discussion above, suggesting the need to disassociate incivility from disagreement. Another possibility is that the targets of the aspersion statements were so general that people did not perceive any possible harm in them. Neither the “justice system” nor various unnamed drivers are likely to take offense from these statements, so those evaluating the statements may not have viewed them as particularly out of line. Future research examining these and other issues will continue to shed needed light on incivility—something that, in its various forms, has considerable implications for democracy.