Journalist Involvement in Comment Sections

Incivility can run rampant in online comment sections. From a democratic angle, incivility on news sites creates reasons for concern. Social science research finds that incivility in the news depresses trust in government institutions.1 Even more, incivility in comment sections can affect readers’ beliefs. Calling this the “nasty effect,” University of Wisconsin Professors Brossard and Scheufele find that uncivil reader comments can change what people think about the news itself.2

From a business angle, some news staff members worry that incivility-laced comment sections can damage their reputation and can harm the overall news brand.3 Although news organizations can employ moderators to remove uncivil comments from these online forums, the practice can be both time-consuming and expensive.

With this in mind, we conducted two field studies with media partners to examine what happens when journalists take a more active role in comment sections. We analyze whether posing questions to commenters can affect incivility and how long site visitors spend on a webpage. We also examine whether the presence of a reporter affects the tenor of the discussion.

Our results demonstrate that a reporter interacting with commenters can improve the civility of the comments. Asking closed-ended questions also promotes greater civility. Finally, we have suggestive evidence that asking questions can increase time on page, but this finding requires additional research. As we summarize below, our results provide real-world guidance for newsrooms.

The Problem

Hope for democratic participation. Pits of partisan incivility. Perceptions of online news comment sections run the gamut.

Optimistically, online comment sections offer a forum for gathering and sharing diverse opinions.4 Indeed, some journalists describe online engagement as necessary for generating a broader discussion about the news.5

Yet commenters presenting carefully-reasoned arguments can run up against angry individuals railing against the government, the news media, and fellow commenters. News comment sections – while hopeful representations of what democratic participation online could be – can turn into uncivil hotbeds of distortion, attack, and vulgarity.6

Can deleterious commenting practices be reversed? Could comment sections be turned into civil centers of citizen engagement? In this report, we present our research on two strategies for intervening in these online spaces.

Key Findings

To Reduce Uncivil Comments:

Ask site visitors a question about the news content with only a few response options. For example:

  • Do you think the new proposal is a good idea or a bad idea?
  • Do you agree with the new bill or not?
  • Should we increase taxes or decrease spending?

Have a reporter visit with people in the comment section, using the following techniques to spark conversation and highlight productive comments:

    • Answer legitimate questions from commenters (e.g. “Good question, Mandy.”)
    • Ask questions of commenters (e.g. “What are your thoughts on that?”)
    • Provide additional information (e.g. “Here’s the link to the bill text.”)
    • Encourage and highlight good discussion (e.g. “Tom, you bring up something interesting.”)

To Increase Time on Page:

Our results are suggestive that asking site visitors a question prior to the comment section can increase time on page, but this result requires more study with a larger sample.

Implications for Newsrooms

Comment sections can seem like the Wild West. They are a space where citizens can express their views, no matter how deplorable or uncivil. Newsrooms have much to gain from these spaces – a loyal online community, feedback on the news, and a reason for visitors to come back and spend more time on the site. But the potential downsides are daunting. Uncivil comment sections can change impressions of the news and may hurt the news brand. How can news organizations foster the benefits of comment sections and limit the downsides? Our research offers two practical suggestions. First, news reporters can get involved in the comment section, engaging politely with site visitors. Although engaging with site visitors in a comment section may seem onerous, the reporter and station in our study did not expend extraordinary efforts to complete this task; the reporter interacted, on average, just over four times.

Second, journalists can direct the conversation by asking questions instead of allowing a free-for-all in the comment section. Closed-ended questions, in particular, seem helpful for inspiring civil interactions. Even more, there is suggestive evidence that posing questions can increase time on page.

The benefits are encouraging. From both a business and a democratic angle, engaging in comment sections has potential.

Political figures from Democratic President Barack Obama and Republican Governor of Texas Rick Perry have called for more civility in politics. As comment sections are one potential source of citizen incivility, this research provides one step toward improving the tenor of the conversation in these venues for citizen engagement.

The Study

First, we look at what happens when newsroom staff engages in comment sections – does having a reporter or a staff member from the newsroom change the substance of the comments? Second, we analyze whether questions posed of commenters affect the quality of the discussion and the time spent on a page. Before turning to the results of our two field research projects, we first provide an overview of what we know about comment sections.

Current State of Comment Sections

News outlets increasingly have integrated comment sections and forums into their sites. In the early 2000s, few news sites – between eight and 30 percent – had comment sections.7 A decade later, 85 percent of online television newsrooms had comment sections.8

Use of news sites’ comment sections varies considerably. In an analysis of just over 100 randomly selected local television news websites, we recorded how many comments appeared in the top five most featured or most popular articles when we arrived on each site. The top five articles on 21 percent of the sites had no comments. Thirty-six percent had 10 or more comments in response to at least one of the top five articles. Although there is great variability at the local level, national news sites, such as CNN, Fox News, or Politico, routinely see thousands of comments on a single story.

Not all citizens participate in comment sections. Yet the percentage of citizens doing so is substantive. The Pew Research Center in 2010, for instance, reported that 32 percent of Internet users had posted a comment on an online news site.9 This study focuses on this active subset of online news users.

In this research, we focus specifically on techniques that could affect the civility of online comments. In particular, we analyze the effects of having newsroom staff engage with online visitors in the comment section and having newsroom staff pose questions prior to a comment section.

Civility in Online Comment Sections

Name-calling. Obscene language. Exaggerating. Each of these behaviors, among many others, exemplifies political incivility – or nasty, offensive, and democratically damaging language that adds little substantive value to a political discussion.10 Instead of being a forum for learning and discovery, comment sections can devolve into a dark cave of name-calling and ad hominem attacks, or as one team of researchers put it, venues of “impoverished and strident discourse.”11

Incivility is present both in politics and in the media. Professor Kathleen Hall Jamieson has been analyzing incivility in Congress over time.12 Her results show variability in levels of incivility between 1935 and 2011. Journalists interviewed as part of the research also noticed periods of greater, and lesser, incivility in Congress.

Incivility also appears in the media. Looking at cable television programs, talk radio shows, political blogs, and columns from syndicated newspaper columnists, Tufts University faculty Sarah Sobieraj and Jeffrey Berry found that nearly 90 percent contained some form of incivility.13 Another study analyzing political news groups in the early 2000s found that 30 percent of the messages were impolite or uncivil.14

Only a few analyses have documented the extent of incivility in comment sections. One study reported that nearly 20 percent of online comments related to nuclear energy and nanotechnology were uncivil.15 According to a recent report from the National Institute for Civil Discourse by University of Arizona faculty Kate Kenski, Kevin Coe, and Steve Rains, over 1 in 5 comments appearing in the Arizona Daily Star during late 2011 were uncivil.16

To be fair, some research suggests that incivility may have some positive business implications. Its presence, at least as part of a news story, may attract an audience.17 Yet the downsides of incivility for news organizations abound.

Incivility negatively affects political attitudes. Exposure to incivility in the news makes people less considerate of opposing viewpoints and less trusting of government.18

Sources ridiculed in comment sections may be less likely to share information in the future. In interviews, reporters worried that if commenters attack their sources, it could be difficult to get information from sources, or get new sources, in the future.19

Uncivil comments may affect the news brand. Interviews with newsroom staff make clear concerns that uncivil comments can affect their “personal and institutional credibility.”20

Uncivil comment sections can affect how people interpret a news story.21 In an experiment conducted by a team of researchers from the University of Wisconsin-Madison, readers of a technology news story who saw only civil comments held the same opinion about the technology as they had before they read the comments. Those who read uncivil comments, however, became more polarized in their opinions of the story, and were more likely to evaluate the technology negatively. In their New York Times editorial describing the study, Brossard and Scheufele note that, “simply including an ad hominem attack in a reader comment was enough to make study participants think the downside of the reported technology was greater than they’d previously thought.” The negative comments held sway in readers’ minds. Such findings reveal the power comments have to affect the interpretation of the news story as a whole.

Since the comment sections of news sites, ideally, are spaces for the public to engage with news, finding ways to increase civility in these forums seems useful.

Increasing Civility in Online Comment Sections

News sites engage in numerous practices to create vibrant, civil comment sections. Some news organizations have internal staffs pre-moderating comments before they appear on the site. Others outsource the task to firms specializing in comment moderation. Others still have experimented with requiring site registration and real names to deter commenting beyond the pale.

News organizations also have come up with a series of rewards and punishments to encourage thoughtful commenting.

In terms of punishments, many sites have some way for people to report inappropriate content. In our analysis of over 100 local television news websites, we found that 70 percent of sites with comment sections had a way for users to flag inappropriate comments.22  This user feedback can be used to identify comments that should be removed. The Orlando Sentinel, for example, monitors comments flagged by Facebook site visitors.23 Sites requiring commenters to log in can block those with a habit of posting unsuitable content.

In terms of rewards, sites have adopted ways of rewarding and highlighting strong comments. The Wall Street Journal, for example, has a list of highly recommended comments and commenters. The Huffington Post awards commenters with Badges for thoughtful commenting and correctly identifying abusive comments.

Our research investigates another strategy for encouraging substantive discourse within comment sections. In this research report, we analyze whether journalist involvement in comment sections affects civility.

Journalist Involvement in Comment Sections

Journalists often read the comments, but rarely interact. In a survey of newspaper reporters in 2010, 98 percent said that they read their outlet’s online comments and nearly 50 percent reported reading them often.24 Yet few were interacting with readers in the comment section; “80 percent of reporters who responded said they ‘rarely’ or ‘never’ respond to online reader comments.”25 Different opinions arise about the role that journalists should play in comment sections.

Some are skeptical about getting involved.

Comments are a reserved space for users. A series of one-on-one interviews with editors revealed that some “thought that the comments were the purview of the users of the site and that the newsroom staff should not respond.”26 Similarly, some journalists reported in interviews that their role is presenting information to the public, not necessarily conversing with news audiences.27

Not enough time to engage properly. Another concern is that it takes too much time to interact in comment sections. In-depth interviews with journalists and editors revealed worries that engaging with audiences in online comment sections would take reporters away from their primary tasks of investigating, writing, and filing stories.28

Targets of rude remarks. Being a target of rude remarks and criticism by commenters can hurt reporter and newsroom morale, leading some to avoid these spaces.29

Others see the merit of engaging site visitors.

Journalists and audiences learn more from each other. Some reporters, when interviewed, described learning from and about their readers from the comment sections.30 In some cases, reporters are also finding new sources among those who are commenting on their stories.31 When interacting, journalists can provide additional information or ask questions, thus furthering the learning process.32

Advancing the brand. News sites can encourage the formation of communities via comment sections in ways that can promote the news brand. Based on interviews with staff at the Washington Post and USA Today, one study concluded that, “participatory journalism was seen as a necessary tool to boost brand loyalty by drawing and retaining audiences.”33

Increasing civility in comment sections. Most important for this research, journalists getting involved in the comment section could enhance civility. First-hand accounts from the Washington Post suggest that journalists can intervene to civilize discourse. Former interactivity editor Jon DeNunzio explains, “I can say from personal experience that when I have gone into threads to explain how our comments work or help users with questions/issues they might have, the tone changes simply because the user realizes someone from The Post is listening.”34 The Washington Post sanctions involvement by newsroom staff in the comment sections. When a journalist interacts, a “WP Staff” insignia denotes the presence of a trusted source.

Strategies of New Staff Involvement in Comment Sections

There are more and less direct ways of getting involved in comment sections, both of which we investigate in our research:

  • Direct involvement would be engaging in the comment section with the commenters.
  • Indirect involvement, however, also could affect the tenor of the conversation. “Have your journalists post topical questions to which readers send written replies.”35

In our study, we analyze whether questions posed of commenters or direct involvement by news staff make any difference to the quality of the discussion or the news organization’s bottom line.

The First Experiment

For our first study, we partnered with a local television news station with a vibrant Facebook community.  The station was an affiliate of a major television network in a top-50 Designated Market Area (DMA). Over 40,000 people had liked the station’s Facebook page.

The study took place between December 2012 and April 2013.36 For each day of the study, the station would post a political story on Facebook and then vary their interactions with commenters according to a randomized schedule.

One of three things happened on each of the 70 political posts associated with the study:

  1. Reporter Interaction: On some days, the station’s well-known political reporter would comment and respond to commenters.
  2. Station Interaction: On other days, the station’s web team, using the station’s logo as their identity, would interact in the comment section.
  3. No Interaction: And on yet other days, no one from the station would interject in the comment section.

We gave the reporter and the station several tips for interacting on the site. Here are examples of how the reporter and station interacted on the site:37

Responding to questions. For example, the reporter posted the following when a commenter asked a question of her fellow commenters: “Good question, Mandy. That would seem to me where legal clashes could happen. I believe the Nebraska bill does include exceptions for the life of the mother.”*

Asking questions. Some of the interactions involved asking questions related to the post topic, such as the following: “We hear from both parties about the issue of undocumented immigrants being freed from federal custody. What are your thoughts on that?”

Providing additional information. When the reporter or station had more information to offer, they were encouraged to add these additional insights to the discussion. For example, the reporter shared this comment and a hyperlink in a discussion related to regulating medical marijuana: “FYI, here’s the text of the recent bill. Just in case you’d enjoy some light reading: (link).”

Encouraging and highlighting good discussion. The reporter and station recognized commenters who were adding to the depth and breadth of the online conversation. In the following example, the reporter acknowledged the contribution of a commenter and then used his comment to ask a question: “Tom, you bring up something interesting: This issue is often seen as ‘Democrats vs. Republicans’ as much as it is seen as ‘Pro-Choice vs. Pro-Life.’ I’m curious, do any of you consider yourselves pro-choice Republicans or pro-life Democrats?”

Analysis of the Posts

We analyzed the posts that were made by the station and then separately analyzed each of the comments left by site visitors. For each of the 70 posts made by the news station, we recorded how user engagement was encouraged.

Conversation Starter. We coded whether the post encouraged people to get involved. Three ways in which posts explicitly invited site visitors to comment included:

  • Close-ended question. The post asked commenters a yes/no or either/or question such as: “The government for the first time is recommending broad new standards to make school snacks healthier. The rules proposed today would ban the sale of almost all high-calorie sports drinks, candy and greasy foods on campus. Good idea or bad idea?” Twenty-six percent of the posts contained this sort of question.
  • Open-ended question. Other posts asked commenters a broad question with many different possible responses: “Several mothers shared heart-wrenching stories at the Capitol today. The women want legislators to pass three controversial bills they hope will prevent more deaths. What are your thoughts on the bills?” Forty-four percent of posts included an open-ended question.
  • Discussion starter. Other posts did not include a specific question, but did have a “discussion starter.” By this we mean that the post invited people to leave a comment without asking a question. For example: “The legislation with perhaps the most potential for debate hasn’t even been filed yet. The bill would limit abortions to 20 weeks, the point at which supporters say the fetus can feel pain. Let us know what you think about this bill in the comments below and feel free to ask any questions you may have!” Thirty-one percent of station posts had a discussion starter.

Across all of the posts, 31 percent did not include an open-ended question, a closed-ended question, or a discussion starter.

Comments. We recorded the number of comments generated by each post. After monitoring the site, we noticed that commenting generally subsided within a day or two after the post was published. Thus, we archived comments three days after publication. On average, posts received 33 comments.

We also coded a host of other attributes about each post, such as the topic of the article, which are described in more detail for interested readers in the appendix.

Analysis of the Comments

Next, we looked at all 2,408 comments that were left on the site in response to the 70 station posts.

To evaluate whether each comment was uncivil, we looked for a series of different characteristics. To develop this list, we drew heavily from research conducted by Sarah Sobieraj and Jeffrey Berry, as well as by Zizi Papacharissi.38 To be coded as uncivil, the post needed to include one or more of the following attributes:

  • Obscene language/vulgarity
  • Insulting language/name-calling
  • Ideologically extreme language (“Sure, Jeffrey, that’s what it was. It couldn’t possibly be that the majority of voters saw through the GOP’s distortions and their desire to only help the rich get richer. Oh, and let’s not forget their plan to kick women back to the 1950’s.”)
  • Stereotyping (“I just don’t want them to smoke & drive. Like drunks do. Let’s make streets safer not worse.”)
  • Exaggerated argument (“HAH! They’ll never do it! These A@$#***les make wayyy too much moola off the public for this stuff.”)

Any comment that contained any of these characteristics was coded as uncivil. Across all of the comments, we coded 44.6% as uncivil.


No Effects on Number of Comments

The number of comments left was unrelated to whether the station or reporter interacted in the comment section. Further, posing an open-ended question, a closed-ended question, a discussion starter, or no question didn’t affect how many comments were left.39

Reducing Incivility

When the reporter interacted in the comment section, incivility declined. As shown in the figure below, the chances of an uncivil comment declined by 15 percent when a reporter interacted in the comment section compared to when no one did so. When the station interacted, it had no effect on the incivility of comments.40

Three explanations may account for the reporter reducing incivility, but not the station. First, the reporter commented more per post than did the station. For posts where the station interacted with commenters, the station had an average of 1.13 comments. When the reporter interacted with commenters, he did so more frequently. On average, the reporter wrote 4.48 comments on posts where he was tasked with interacting. Higher rates of commenting may explain why the reporter, and not the station, reduced incivility.

Second, the station had, from time to time, commented in the past. The novelty of the reporter interacting may account for the results.

Third, it is possible that seeing a recognizable reporter from the news broadcast – as opposed to a generic station logo accompanying each comment – sparked additional civility. This study cannot sort out whether one, two, or all three, of these explanations, account for the effects. What is clear, however, is that having a reporter engage in the comment section can affect the civility of the comments. (See image in full report.)4142

The way in which the post was constructed also mattered for the civility of the comments. Posts that used a closed-ended question, asking respondents a yes/no question for example, yielded more civil comments than posts that did not include any invitations for commenter involvement. Asking a closed-ended question in a post reduced the chances of an uncivil comment by 9 percent compared to posts without any invitation for user interaction. See the figure on the previous page.

Although not depicted in the figure, there was some evidence that discussion starters, posts that prompted commenters to get involved but didn’t specifically ask a question, increased incivility. The finding, however, requires more research.43

The Second Experiment

In the previous study, we analyzed whether a news station should ask questions to affect the civility of online comments. The results showed that asking closed-ended questions was related to more civil comments.

In our second study, we test whether questions matter for a news organization’s bottom line. In this study, we analyzed the effects of adding a question prior to a comment section. The idea was that a question could help to direct the conversation. Here, we analyzed two main outcomes: the number of comments and time on page.

We partnered with CultureMap, a culture and arts news organization based in Texas. CultureMap was an ideal partner because they have different websites for multiple locations that sometimes share content. We worked with CultureMap in Houston and Austin to conduct our second study.

The study worked like this: when CultureMap would post the same story on the Houston and Austin sites, one of the sites also would post a question prior to the comment section, which was included at the bottom of the story. The other site would include the same story, but would not include a question. An example screen capture of the Houston and Austin sites can be found in the Appendix. We followed this procedure 34 times between December 2012 and April 2013.

Here are a few examples of the types of questions that CultureMap included:

  • Do you think Rick Perry is softening his position on homosexuality or is the media misinterpreting his speech? Tell us what you think in the comments section below.
  • Do you think casinos and Tesla’s electric cars belong in Texas? Tell us your thoughts in the comment section below.
  • Why do you think so many centers of higher education are receiving bomb threats? Tell us in the comment section below.


No Effects on Number of Comments

The number of comments left on the Austin and Houston pages during the period of this study varied between 0 and 30. There were no differences in the number of comments based on whether a question appeared on the site, however.44

Increasing Time on Page

There were indications that the average time on page increased when the question was included. On average, people spent 2 minutes, 43 seconds on the pages with a question and 2 minutes, 26 seconds on pages without a question.45 In general, however, the difference was not statistically significant.

More research with a larger sample is required to analyze whether time on page is affected by the presence of a question, but this result is suggestive and consistent with our suspicions.


In addition to the attributes mentioned in the main body of the text, we also coded for several other attributes of the posts and comments which were controlled in our analysis. We explain each below.

Post Attributes

Topic. Posts discussed crime/guns (26%), economy (26%), education (30%), and health (23%). Posts could be coded for multiple categories; for example, a post about how state budget cuts might affect education would fall under the economy topic and the education topic.

Conversation Starter. As detailed in the text, we recorded whether each post included a closed-ended question (26%), an open-ended question (44%), or contained a discussion starter where commenters were prompted to participate without a question (e.g. “Leave your comments below”; 31%). Thirty-one percent of posts did not include an open-ended question, a closed-ended question, or a discussion starter.

Comment Attributes

Partisanship. If the comment mentioned a political party (e.g. Democrat or Republican) or political ideology (e.g. liberal or conservative), we noted it. Five percent of comments included this language.

Reporter/Station Comment. We recorded whether the comment made was by the reporter, the station, or an audience member. Only 5 percent of posts were from the station or reporter. These were removed from our analyses.

Incivility. Comments were coded as “uncivil” if they contained any one of the following attributes: (1) Obscene language/vulgarity, (2) insulting language/name-calling, (3) ideologically extreme language, (4) stereotyping, or (5) exaggerated arguments, over-the-top or overgeneralized arguments. Across all of the comments, we coded 44.6% as uncivil.

Reliability of Coding

We went through several rounds of coding to establish reliability. In this process, multiple people code the same post or comment, and then their responses are compared to see if they reached the same conclusion about the content. For example, we measure whether multiple people identify incivility in a comment (or not) when they follow the rules we established for identifying incivility in a comment. To measure reliability, we used Krippendorff’s alpha. Values of Krippendorff’s alpha closer to 1.00 indicate more reliable coding, with values in excess of .80 considered particularly strong and values in excess of .67 considered acceptable.



Stroud, Natalie Jomini, Muddiman, Ashley, Scacco, Joshua, Curry, Alex, and Peacock, Cynthia. (2014, September). Journalist Involvement in Comment Sections. Center for Media Engagement.

  1. Mutz, D. C., & Reeves, B. (2005). The new videomalaise: Effects of televised incivility on political trust. American Political Science Review, 99, 1-15.
  2. Brossard, D., & Scheufele, D. (2013, March 2). This story stinks. New York Times. Retrieved from; Anderson, A. A., Brossard, D., Scheufele, D. A., Xenos, M. A., & Ladwig, P. (forthcoming). The “nasty effect:” Online incivility and risk perceptions of emerging technologies. Journal of Computer-Mediated Communication.
  3. Singer, J. B., & Ashman, I. (2009). “Comment is free, but facts are sacred:” User-generated content and ethical constructs at the Guardian. Journal of Mass Media Ethics, 24, 3-21. See also Hermida, A., & Thurman, N. (2008). A clash of cultures: The integration of user-generated content within professional journalistic frameworks at British newspaper websites. Journalism Practice, 2, 343-356.
  4. Bimber, B. (1998). The internet and political transformation: Populism, community, and accelerated pluralism. Polity, 31, 133-160.
  5. Hermida, A., Domingo, D., Heinonen, A. A., Paulussen, S., Quandt, T., Reich, Z., Singer, J. B., & Vujnovic, M. (2011, April). The active recipient: Participatory journalism through the lens of the Dewey- Lippmann debate. Paper presented at the International Symposium on Online Journalism, Austin, Texas.
  6. Diakopoulos, N., & Naaman, M. (2011, March). Towards quality discourse in online news comments. In Proceedings of the ACM 2011 conference on Computer Supported Cooperative Work (pp. 133- 142).
  7. Chan-Olmsted, S. M., & Suk Park, J. (2000). From on-air to online world: Examining the content and structures of broadcast TV stations’ web sites. Journalism & Mass Communication Quarterly, 77, 321- 339; Dibean, W., & Garrison, B. (2001). How six online newspapers use web technologies. Newspaper Research Journal, 22, 79-93; Schultz, T. (1999). Interactive options in online journalism: A content analysis of 100 U.S. newspapers. Journal of Computer-Mediated Communication, 5. doi: 10.1111/ j.1083-6101.1999.tb00331.x
  8. Stroud, N. J., Muddiman, A., & Scacco, J. M. (forthcoming). Engaging audiences via online news sites. In H. Gil de Zuniga (Ed.), New agendas in communication: New technologies and civic engagement. New York, NY: Routledge.
  9. Pew Research Center (2010). Trend Data (Adults). Retrieved from Static-Pages/Trend-Data-%28Adults%29/Online-Activities-Total.aspx.
  10. Ben-Porath, E. N. (2010). Interview effects: Theory and evidence for the impact of televised political interviews on viewer attitudes. Communication Theory, 20, 323-347; Brooks, D. J., & Geer, J. G. (2007). Beyond negativity: The effects of incivility on the electorate. American Journal of Political Science, 51, 1-16; Fridkin, K. L., & Kenney, P. J. (2008). The dimensions of negative messages. American Politics Research, 36, 694-723; Mutz, D. C. (2007). Effects of “in-your-face” television discourse on perceptions of a legitimate opposition. American Political Science Review, 101, 621-635; Papacharissi, Z. (2004). Democracy online: Civility, politeness, and the democratic potential of online political discussion groups. New Media & Society, 6, 259-283.
  11. Cenite, M., & Zhang, Y. (2010). Recommendations for hosting audience comments based on discourse ethics. Journal of Mass Media Ethics, 25, 293-309.
  12. Jamieson, K. H. (1997, March 1). Civility in the House of Representatives. APPC Report #10. The Annenberg Public Policy Center at the University of Pennsylvania. Available at:; Annenberg Public Policy Center Report (2011, September 28). Civility in Congress (1935-2011) as reflected in the Taking Down process. APPC Report #2011-1. The Annenberg Public Policy Center at the University of Pennsylvania.
  13. Sobieraj, S., & Berry, J. M. (2011). From incivility to outrage: Political discourse in blogs, talk radio, and cable news. Political Communication, 28, 19-41.
  14. Papacharissi, 2004.
  15. Anderson, A. A., Xenos, M. A., Brossard, D., & Scheufele, D. A. (2012, August). Caustic comments: Measuring incivility in online comments and testing its effects on political participation. Paper presented at the 2012 Annual Conference of the Association for Education in Journalism and Mass Communications, Chicago, IL.
  16. Kenski, K., Coe, K., & Rains, S. (2012, October 13). Patterns and determinants of civility in online discussions. Report prepared for the National Institute for Civil Discourse.
  17. Mutz & Reeves, 2005.
  18. Borah, P. (forthcoming). Does it matter where you read the news story? Interaction of incivility and news frames in the political blogosphere. Communication Research; Mutz, 2005; 2007.
  19. Diakopoulos & Naaman, 2011; Loke, J. (2012). Old turf, new neighbors: Journalists’ perspectives on their new shared space. Journalism Practice, 6, 233-249.
  20. Hermida & Thurman, 2008; Singer & Ashman, 2009.
  21. Anderson, et al., forthcoming; Brossard, & Scheufele, 2013.
  22. Stroud et al., forthcoming.
  23. Tenore, M. J. (2012, April 12). How 5 news sites monitor, moderate conversation about Trayvon Martin stories.
  24. Santana, A. D. (2011). Online readers’ comments represent new opinion pipeline. Newspaper Research Journal, 32, 66-81.
  25. Santana, 2011, p. 73.
  26. Diakopoulos & Naaman, 2011, p. 140.
  27. Hermida et al., 2011.
  28. Paulussen, S., & Ugille, P. (2008). User-generated content in the newsroom: Professional and organizational constraints on participatory journalism. Westminster Papers in Communication and Culture, 5, 24-41; Singer, & Ashman, 2009.
  29. Singer & Ashman, 2009.
  30. Bowman, S., & Willis, C. (2003, July). We media: How audiences are shaping the future of news and information. Retrieved from Singer & Ashman, 2009.
  31. Lysak, S., Cremedas, M., & Wolf, J. (2012). Facebook and Twitter in the newsroom: How and why local television news is getting social with viewers? Electronic Media, 6, 187-207.
  32. Diakopoulos & Naaman, 2011.
  33. Vujnovic, M., Singer, J. B., Paulussen, S., Heinonen, A., Reich, Z., Quandt, T., Hermida, A., & Domingo, D. (2009). Exploring the political-economic factors of participatory journalism: A first look into self-reports by online journalists and editors in ten countries. Journalism Practice, 4, 290.
  34. Phelps, A. (2012, February 1). The Washington Post tries a new weapon to fight the trolls: Humans. Retrieved from fight-the-trolls-humans/.
  35. Hermida & Thurman, 2008, p. 345.
  36. Note that we did not conduct the study on weekends, holidays, or days when there was breaking news that prevented the station from taking part.
  37. As we discuss later, the reporter interacted more and used these techniques more frequently than the station.
  38. Papacharissi, 2004; Sobieraj & Berry, 2011.
  39. As the dependent variable (number of comments) was measured at the post level, rather than the comment level, ANOVA was used. An ANOVA predicting the number of comments revealed no significant effect of interaction (F(2,64)=.57, p=.57), open-ended questions (F(1,64)=.18, p=.67), closed-ended questions (F(1,64)=.77, p=.38), or discussion starter (F(1, 64)=2.52, p=.12). Note that we removed all comments made by the reporter or the station from the analysis as the presence of these comments could inflate artificially the number of comments.
  40. We also analyzed whether incivility declined over the course of the study. We included a time variable in the analysis, numbered consecutively for each post as part of the study. The variable was not significant when added to the incivility model (B = -.002, SE=.004, p=.52).
  41. This study is appropriately understood as a quasi-experiment. Although we randomly assigned the type of interaction (e.g. whether the station interacted, a reporter interacted, or neither), other factors, such as the topic of the post, were not under our control. Further, we did not randomly assign the conversation starter (e.g. open or closed-ended question). For this reason, we control for post topic, retaining education and health in our final model as these were the only two significant topical predictors. We also control for mentions of partisanship/ideology in the comments. Results are from an HLM analysis with a random intercept for each post. In the analysis, incivility on station interaction days was not significantly different from no interaction days (B=.09, SE=.19, p=.46). Incivility on reporter interaction days was significantly different from no interaction days (B= -.63, SE=.32, p<.05). To create the figure, control variables were held constant at mean and modal values. We examined one other indicator of incivility, shutting down the conversation (e.g. “It would help many Republican leaders if they knew when to close their mouths and open their minds.”). When this also is included as an indicator of incivility, incivility on reporter interaction days drops to marginal significance in comparison to no interaction days (B= -.57, SE=.34, p<.10).
  42. Results are from the same analysis reported in the “Probability of an Uncivil Comment by Interaction” figure. Here, an open-ended question did not produce any different level of incivility compared to no conversation starter (B=-.13, SE=.17, p=.46). A closed-ended question did, however, yield less incivility than no conversation starter (B=-.37, SE=.19, p<.05).
  43. Discussion starters were marginally significant predictors of higher levels of incivility (B=.48, SE=.28, p<.10).
  44. The raw data are presented in text and the difference between the pages with and without questions is not significant using raw data (t(33)=.36, p=.72). As detailed in the next footnote, we also standardized the data and examined differences using other statistical tests. For the number of comments, none of these tests were statistically significant.
  45. The raw data are presented in text and the difference between the pages with and without questions are not significant using raw data (t(33)=1.06, p=.30). Yet the raw data do not take into account differences between the Houston and Austin sites. To take this into account, we first standardized the data from Houston and Austin separately. With these data, the differences are statistically significant. Site visitors spent more time on pages with a question compared to pages without a question (t(33)=2.18, p<.05). This difference holds when using the non-parametric Wilcoxon signed rank test (Z=2.14, p<.05). It also holds if we use repeated measures ANOVA controlling for (a) city where the question appeared and (b) whether the headlines differed (in 50% of the cases, there were small headline differences between the sites). Here, the paired comparison between the pages with and without questions was significant (mean difference = .47, SE=.22, p<.05). In 18 percent of the cases, the pages were posted on the Houston and Austin sites on different days. The results were identical across all of the tests mentioned above if these cases are excluded. If we use a 10% Winsorization of the data, where the highest and lowest 10% of values are replaced with the next highest and lowest value, however, the data again are not significant, although the means again trend in the same direction (Page with question M=2.53, SE=.11; Page without question M=2.44, SE=.11, t(33)=.65, p=.52). As the significance of the findings is dependent on the statistical test employed, we urge caution in applying this result. Although not always significant, we are encouraged that the direction of the result is consistent – the question page yields more time than a page without a question.