TY - JOUR
AU - Quiring,, Oliver
AB - Abstract While the social, political, and journalistic relevance of user comments on online news items has been discussed intensively, no study has tried to examine why some online news discussions are more interactive than others. Based on the rationale of news value theory, this study argues that so-called discussion factors in user comments indicate general relevance to later users to respond to them. Qualitative interviews with users who comment on news stories online and a quantitative content analysis of 1,580 user comments showed that the discussion factors uncertainty, controversy, comprehensibility, negativity, and personalization can explain interactivity in news discussions. Further, different technological implementations of the comment function seem to have a limited influence on the effects of these discussion factors. Many news media allow users to comment on the news stories published either on their websites or on related presences on social network sites (Diakopoulos & Naaman, 2011; Reich, 2011; Ruiz et al., 2011). These comments usually appear below the news items. In recent years, user comments have evolved into a standard feature of online news (Walther, & Jang, 2012; Weber, 2013). They are often considered as the most popular form of public online participation (e.g., Reich, 2011; Weber, 2013). According to Reich (2011), news stories without user comments “are becoming rare and starting to look awkward, even suspicious” (p. 97). With regard to actual participation, a survey of the Pew Research Center found that 25% of adult U.S. Internet users have already commented on online news or news blogs at least once, and that 37% appreciated the comment function (Purcell, Rainie, Mitchell, Rosenstiel, & Olmstead, 2010). Owing to their large reach and their potential effects on the audiences' perceptions of online news, it is important to explore the dynamics of user comments (e.g., Anderson, Brossard, Scheufele, Xenos, & Ladwig, 2013; Diakopoulos & Naaman, 2011; Lee & Jang, 2010; Reich, 2011; Ruiz et al., 2011; Singer, 2009; Weber, 2013). For example, various characteristics of news articles have been found to affect the total number of comments posted (Weber, 2013). Studies have also reported that between 20 and 50% of the comments analyzed contain responses to the comments posted by other users (e.g., Ruiz et al., 2011; Singer, 2009). Such user-to-user interactions have been theorized as desirable from both a journalistic and a deliberation perspective because they could contribute to shaping a democratically valuable and vivid interpersonal discourse on topics of public interest (Boczkowski & Mitchelstein, 2012; Freelon, 2010; Ruiz et al., 2011). However, little is known about why users respond to the comments of other users. Research on the effects of user comments has primarily analyzed cognitive responses of users who are exposed to other users' comments (e.g., Anderson et al., 2013; Lee & Jang, 2010).1 Moreover, studies analyzing posting behavior have largely neglected the influence of previous comments (e.g., Boczkowski & Mitchelstein, 2012; Weber, 2013). Hence, our exploratory investigation aims at identifying so-called discussion factors in user comments that help to explain why certain comments receive response comments (i.e., feedback2) from other users. As a theoretical starting point, we use news value theory, which states that journalists and media users select news items depending on news factors such as controversy, negativity, and personalization (Eilders, 2006; Galtung & Ruge, 1965). We assume that users reading comments behave similarly and respond in particular to comments that include specific discussion factors. Such user comments are more likely to receive response comments than comments that do not include these specific factors. However, research on news value theory has shown that secondary factors—such as the visibility of news items—moderate the influence of news factors on the news selection of media users (Eilders, 2006). Thus, we have assumed that the frequency of user-to-user-interactions also depends on the technical implementation of the comment function. In the following sections, we first provide a brief summary of our understanding of interactivity in online (news) discussions before outlining the theoretical approach that leads from news value theory to our idea of discussion factors. We then describe our two-step empirical approach to (a) identify discussion factors in Study 1 (explorative) and (b) test their effects on the behavior of users of different news platforms in Study 2 (confirmative). Interactivity in online (news) discussions We define online user comments as a subcategory of media-stimulated interpersonal communication that is published directly below news items on news websites or on news media presences within other online communication services (Ziegele & Quiring, 2013). Further, an online news discussion is defined as a sequence of user comments on a particular news item (cf. Ruiz et al., 2011). Because different scholars pay attention to different aspects of interactivity (cf. Quiring, 2009; Rafaeli & Ariel, 2007), the concept of interactivity can be defined from three different perspectives: (1) as an attribute of technical media systems (e.g., Sundar, 2004), (2) as an attribute of the communication process (e.g., Rafaeli & Sudweeks, 1997), and (3) as an attribute of the perceptions of users (e.g., Quiring, 2009). Concerning perspective 1 (structure), studies have found that the technical/structural attributes of online services (CMC settings) determine to a certain degree the number of interactions and guide the way in which they take place (e.g., Jones & Rafaeli, 2000; Preece, 2001; Wright & Street, 2007). In this context, Jones and Rafaeli (2000) have coined the term “discourse architecture,” which describes “the way in which a virtual public's technology structures discourse” (p. 217). User comment sections vary with regard to several features, for example, with regard to how many comments are displayed per page and how these comments are sorted. In addition, the providers of comment sections apply different policies with regard to how extensively they moderate discussions or whether they require users to register, among other things (e.g., Reich, 2011; Ruiz et al., 2011). Within the news value framework, different discourse architectures can be seen as secondary factors that have been found to affect the news selection of media users in traditional news value research (Eilders, 2006; Weber, 2013). Thus, when analyzing interaction structures in user comments, it seems important to consider the respective platform's discourse architecture as a moderating influence. Concerning perspective 2 (process), scholars often distinguish between interactions between users and systems on one hand and between users themselves on the other hand (e.g., Bucy, 2004; Rafaeli & Sudweeks, 1997). We have concentrated on user-to-user interactions and have looked at the degree to which different user messages are interconnected. We assume messages to be interconnected when they address at least one previously posted user message, that is, contain reciprocal communication (Walther, & Jang, 2012). Hence, user-to-user-interactions occur when a user writes a response comment to a message previously posted by another user. Both perspectives (structure and process) are linked by the perceptions of users (perspective 3). That is, users have to perceive the interactive potential offered by a CMC setting (perspective 1) to make full use of the technical structure and enter interactions with other users (perspective 2). Theoretical approach: From news factors to discussion factors Although user-to-user interactions are highly prevalent in some news discussions (Ruiz et al., 2011; Singer, 2009), little is known about the reasons why users interact by writing response comments to previous posts. In the following text, we hypothesize that specific message characteristics in user comments stimulate response comments from subsequent users. These message characteristics are referred to as discussion factors because responding to a user comment can be considered as the first necessary step to initiate an interpersonal discussion. News value theory provides a promising starting point for identifying the discussion factors in user comments. The roots of this theory can be found in journalism studies and in journalists' handbooks (e.g., Lippmann, 1922; Warren, 1934). According to news value theory, journalists select events for news production that contain so-called news factors (Galtung & Ruge, 1965). These news factors vary according to different authors, but most of them analyze an event's proximity (geographical, cultural, or economic distance between an event and the country in which it is reported), continuity (duration of news coverage), prominence/influence (publicity and/or power of persons, groups, or institutions), personalization (portrayal of single persons), reach (number of persons affected), unexpectedness (unpredictability), damage (negative consequences), and controversy (portrayal of disagreement) (for exhaustive taxonomies cf. Galtung & Ruge, 1965; Schulz, 1982). Taken together, these news factors establish the news value of an event, which in turn determines its probability of being reported (Galtung & Ruge, 1965). Beyond its original focus on journalists (for a review, see Staab, 1990), research on news value theory has shown that media users apply similar selection criteria (a) when choosing between different news items, and (b) when talking about them in interpersonal communication (Deutschmann & Danielson, 1960; Eilders, 2006; Galtung & Ruge, 1965; Shoemaker & Cohen, 2006). As a consequence, news factors have been discussed as universal relevance3 indicators (Eilders, 2006; Galtung & Ruge, 1965; Weber, 2013) that trigger both cognitive and behavioral responses of journalists and media users. Arguments supporting this claim have been derived from evolutionary theory, general perception psychology, and social psychology. These different perspectives share the common assumption that news factors activate pre-existing cognitive schemes and thereby facilitate the process of recognizing and categorizing important incidents. For example, it can be argued that individuals automatically assign relevance to the news factors damage and unexpectedness because they have learned in the course of evolution that these factors constitute events that might threaten their life or well-being (Eilders, 2006; Shoemaker, 1996). Other factors—such as prominence or controversy—generate relevance primarily by suggesting social significance (Shoemaker & Cohen, 2006). That is, these factors activate cognitive schemes related to an individual's position in the society or his/her shared norms and values (Eilders, 2006; Weber, 2013). Finally, messages indicative of news factors such as continuity can be processed more easily because they activate schemes that are based on prior (factual) knowledge (Eilders, 2006). Of course, these explanations are not independent from each other (Shoemaker & Cohen, 2006). All of them suggest that news factors can increase a user's involvement. High involvement, in turn, can be linked to an increased willingness to express one's opinion (Diakopoulos & Naaman, 2011; Weber, 2013). On the basis of news value theory, recent research has begun to analyze links between news characteristics and participatory online behavior (e.g., Weber, 2013; Ziegele & Quiring, 2013). Ziegele and Quiring (2013) have suggested a model of online discussion value which posits that message factors from both news items and previously posted user comments affect the involvement of users and ultimately increase their willingness to engage in online news discussions. Focusing on news factors in news items, Weber (2013) reported that some factors (i.e., impact, frequency, facticity, and proximity) significantly affect the aggregate number of comments posted on news items. However, between 20 and 50% of the comments do not respond to the journalistic news article but rather to a previously posted user comment (Ruiz et al., 2011; Singer, 2009). These user-to-user-interactions—which we define in terms of news value theory as a behavioral response to a message stimulus—cannot be explained by news factors in news items alone (Ziegele & Quiring, 2013). Rather, it can be hypothesized that user comments themselves contain factors that, in part, may resemble news factors and that explain an additional share of the interactivity in online news discussions. Transferring the aforementioned arguments of news value theory to user comments, it can be assumed that users preferably read high-involving comments that can be processed easily and efficiently. Such comments should include factors that affect personal well-being (evolutionary argument), challenge personal identity within a broader societal context (socialization argument), and/or activate prior (factual) knowledge (cognitive argument). Ultimately, comments including such factors may not only increase the probability of generating a cognitive response (i.e., paying attention to the respective comments), but they may also stimulate a behavioral response (i.e., replying to the respective comments). Previous studies on online participation only provide hints on what discussion factors might look like. For example, various studies have investigated the prevalence and users' perception of conflict, incivility, and impoliteness in public discussions (e.g., Ng & Detenber, 2005; Papacharissi, 2004; Stromer-Galley, 2003). Such variables can be seen as interpersonal indicators for the news factor controversy (Eilders, 2006) or aggression (Schulz, 1982). Research on online deliberation has emphasized the significance of well-argued contributions in discussion processes (e.g., Dahlberg, 2001; Ruiz et al., 2011). Using arguments can be seen as an indicator of the news factor facticity (Eilders, 2006). Yet, in both cases, little is known about whether these factors stimulate or thwart interactivity in online news discussions. For example, uncivil and impolite comments might challenge a user's shared norms and ideals and thereby implicate social significance as the factor that resulted in a response to the comment (cf. Herring, Job-Sluder, Scheckler, & Barab, 2002). On the other hand, these comments have also been discussed as inhibitors of participation (e.g., Ng & Detenber, 2005; Papacharissi, 2004). To summarize, news value theory provides helpful insights into the possible nature and psychology of discussion factors in user comments. However, it remains unclear as to whether the crucial factors in user comments are congruent with the news factors that have been found to be valid for the selection of news items by both journalists and media users (conflict, controversy, continuity, reach, and unexpectedness, cf. Eilders, 2006), and whether their effects depend on different technological implementations of the comment sections. Thus, we think that a two-step empirical approach is necessary to identify discussion factors in user comments. In Study 1, we conduct an in-depth analysis of how users perceive and evaluate online news discussions and the comments of other users. The characteristics of the user comments that these users perceive as worth responding to will be discussed in light of news value theory and they will be converted into a catalog of discussion factors. The influence of these discussion factors on the probability of a response to a user comment will then be tested for different technical platforms in Study 2. Study 1 This study's primary aim was to identify discussion factors by exploring when and why users respond to other users' comments (RQ1). The study also aimed at exploring between-platform differences in users' perceptions and evaluations of online discussions (RQ2). Procedure During the fall of 2011, 25 respondents from all over Germany agreed to participate in guided face-to-face interviews. We used purposive sampling to include a broad variety of “typical” users who comment on news items (Tong, Sainsbury, & Craig, 2007). The sampling criteria were frequency of commenting (occasionally, regularly4), age (up to 30 years, 30 years and older), sex (male, female), and preferred news sources (“traditional” news websites and news websites on Facebook). The procedure was as follows: Initially, 30 students of a master's level class were asked to recruit participants in their social networks. The respondents were categorized into the sampling matrix. Then, “missing cells” of the sampling matrix were filled up by directly writing to commenting users on various news websites and on the news companies' Facebook sites. Overall, we interviewed 8 female and 17 male participants. The gender distribution of the sample reflects previously reported findings that people who comment on the news are predominantly male (e.g., Springer & Pfaffinger, 2012). The sample covered a wide variety of ages (from 20 to 55 years, M = 30 years) and different commenting behaviors: Twelve participants commented on news items on a regular basis while 13 did so occasionally. Furthermore, the participants posted comments on different platforms, and eight participants had already discussed news items both on news websites and on Facebook. Finally, various educational backgrounds were represented in the sample. The interview procedure followed the guidelines and recommendations of several methodological handbooks (e.g., Lindlof & Taylor, 2002). The interviews were conducted at different places outside the university, including the homes of the participants and public facilities (e.g., cafés). The semistructured interview guidelines contained open-ended questions on the characteristics of the news items the participants commented on as well as on the perceived quality of online news discussions, the characteristics of useful comments, and the reasons for either responding to other users' comments or ignoring them. At the beginning of each interview, the participants were asked to reconstruct the most recent situation in which they had published a user comment. This narrative was followed by the question “Please tell me what the comments of other users mean to you.” Additional questions were asked about the characteristics of the comments that the participants perceived to be worth responding to. These questions included “Can you describe the comments you respond to?” and “How would you characterize your own comments that received a response from other users?” At the end of each interview, the interviewer handed out a printout of an online news discussion in which the respective interviewee had participated. Then, the interviewee was asked to identify the characteristics of the news item and the other comments that drove his or her decision to participate. Although this kind of retrospective reasoning might not be adequate to detect the exact reasons for posting a comment, it nevertheless tells a lot about the subjective nature of commenting. These responses served as the basic material to identify discussion factors. However, only factors that were mentioned by at least three participants were considered in our analysis. Five interviewers were trained regarding the theory and practice of performing semistructured interviews as well as the details of the guidelines. Prior to the interviews, the participants were asked to sign an informed consent form. Interviews with the participants lasted between 20 and 90 minutes, and the findings were anonymized and transcribed afterwards for further analysis. Theoretical coding (Strauss & Corbin, 1990) was chosen to systematically screen the material for potential discussion factors. Results Overall, the users we interviewed regarded other users' comments as an essential component of online news stories. Not every comment, however, was perceived as equally valuable. The participants reported various factors that increase or decrease the overall discussion value of user comments. These factors can, essentially, be classified within the framework of the original news value theory. However, they have to be conceptualized from a media user's perspective in order to serve as discussion factors (RQ1). Interestingly, the occurrence of specific discussion factors was not always associated with uniform behavior of the users we interviewed: While some users felt engaged by certain discussion factors and indicated that they would reply to comments including these factors, other users asserted that they would never respond to comments containing the same factors. The heterogeneity of the responses indicates a complex interaction between the users' personality, their situational motives, and the message characteristics as already theorized in the model of the online discussion value (Ziegele & Quiring, 2013). Analyzing the direction of influence of the discussion factors will be the subject of Study 2. In Study 1, we identified the following discussion factors: Aggression In news items, aggression occurs when an actor threatens others verbally or by armed force (Staab, 1990). In online news discussions, the users we interviewed described comments as aggressive when the authors “shouted,” when they insulted other users or nonpresent others (e. g., a politician or other persons described in the news item), or when they accused other users or nonpresent others of being incompetent [ex1]. [ex1] When you notice that the emotion turns into an aggression, that is, when people are insulted in discussions about the chancellor and cuss words are used (…) then I say, ok, this does not make any sense and must be ignored. (m, 24, occ.5) While most of the users we interviewed agreed that they would not respond to such aggressive comments, others felt challenged to rebuke the authors of these postings. Controversy The original news factor controversy describes the portrayal of dissent in news items (Eilders, 2006). With regard to comments, the users we interviewed perceived controversy to stem from a wide range of factors causing dissent, including stereotypes, polemical statements, unfounded demands, provocative questions, and exaggerations [ex2]. [ex2] When someone conjures up (…) the total meltdown of the financial system, then (…) I'm almost curious about talking him out of this because this is a little too exaggerated and big. (m, 25, occ.) Unlike aggression, the participants perceived that controversial statements explicitly referred to the topic under discussion. Some of the interviewees welcomed such provocative statements as an opportunity to join the discussion. Others, however, were afraid of publicly opposing comments that contradicted their own views and beliefs. Facticity In news stories, high facticity occurs when an article reports concrete actions and events, while stories that evaluate, interpret, and analyze situations score low on facticity (Weber, 2013). Similarly, the users we interviewed considered “objective” and factual comments that show a high level of expertise or include proved additional factual knowledge to have high facticity [ex3]. [ex3] Some comments are really useful, for example, when someone knows more than the journalist and (…) posts a link and says ‘look, it's like that!’ (m, 27, reg.) Some interviewees preferred responding to factual comments but others perceived that—due to their high level of objectiveness—such comments leave no room for discussion. Unexpectedness The news factor unexpectedness describes events that are rare and/or contradict existing expectations (Galtung & Ruge, 1965). In news discussions, the participants of the current study reported being “surprised” by comments that include topic drifts or a different perspective of the issue discussed [ex4]. [ex4] Sometimes, these comments contain surprisingly great new ideas which you had not considered for yourself. (m, 55, occ.) Most participants reported that such unexpected comments attracted their attention and—in the case of perspective changes—extended their understanding of the issue under discussion. Regarding their response behavior, some interviewees perceived that these comments “supplied” discussions with new aspects that could be talked about. However, other participants did not feel a need to discuss topic drifts or to appreciate or criticize new perspectives. Negativity Various news value studies have argued that journalists and audiences preferably select negative news items because negative messages are deviant and represent potential threats (e.g., Galtung & Ruge, 1965; Shoemaker & Cohen, 2006). In news discussions, the participants perceived that negativity primarily appeared in comments that universally dismissed a news story's content or a previous user comment and thereby emanated a negative overall tone [ex5]. [ex5] Especially in the case of political news items, most comments concentrate on bashing everything, very much like you know it from the yellow press. (m, 20, reg.) With regard to their response behavior, our participants overwhelmingly preferred comments that were written in a positive and inviting tone. They perceived that responding to negative comments might not stimulate meaningful discussions but rather prompt the authors of these comments to dismiss a user's response across-the-board as well. Personalization Personalized news items portray single persons or report examples of individual experience (Weber, 2013). With regard to comments, the participants ascribed a “personal touch” to postings that illustrate how a news story affects the author personally (e.g., comments including an exemplar about a user's personal experience) or that explicitly address single or multiple other users to explain why the comment is relevant to them [ex6]. [ex6] (…) but when I address someone directly, for example, and then comment on his posting, then I usually get a response. (m, 20, reg.) Again, the effects of personalization on the participants' response behavior were ambiguous. Some users preferred responding to personalized comments, but others were afraid of getting publicly entangled in personal debates and therefore abstained from replying to such comments. Simplification According to Östgaard (1965), journalists intentionally reduce the complexity of issues when reporting them in order to make them more easily understandable and to attract more readers. The participants of the current study reported the authors of comments to behave similarly—although often unintentionally—by simplifying the causes of issues or by blanking out complex issue backgrounds [ex7]. [ex7] Frequently, users claim that they know exactly why something has happened. But their comment reveals that they neither have read the entire news story to understand the whole issue, nor did they obviously ever inform themselves about its actual complexity. (m, 26, occ.) Regarding their response behavior, some participants felt the need to inform the authors of these simplified comments about the actual complexity of the issue under discussion. Others, however, anticipated that such a behavior would result in an exhausting and unwanted discussion. Comprehensibility Although it is not considered as a genuine news factor, Östgaard (1965) refers to comprehensibility as the journalistic effort “to use simple words and sentences in order to get the ‘message across’ more easily” (p. 45). In comments, the interviewees primarily paid attention to characteristics that decrease comprehensibility. More specifically, they argued that the use of metaphors and irony [ex8] made it difficult to grasp the meaning of comments. [ex8] Some things get lost in written communication (…). It's always difficult to express irony in your comments in a way that others can understand it. (m, 30, occ.) Few participants reported that they ask the authors of such inscrutable comments to clarify their argument. Most of the participants who mentioned this factor, however, asserted that they ignored incomprehensible comments. Uncertainty Early news value research argued that journalists preferably try to provide their audience with facts and answers and that they therefore select events that are free from uncertainties (Galtung & Ruge, 1965). In the current study, however, many users reported that they actively create uncertainty in their comments by asking questions about additional facts or the possible meaning of news items [ex9]. [ex9] Rather than telling my own opinion, I'm asking questions when I think that something has remained unclear. And I want these questions to be answered, and (…) when another user can answer them, then I'm fine. (f, 21, occ.) This strategy was generally perceived as a powerful measure to receive feedback. Participants who encountered questions felt committed or invited to respond to them, either because they were challenged or because they wanted to demonstrate their knowledge. Humor According to the perception of the interviewees, many news items are not discussed entirely seriously. Instead, across all news reports, the participants reported that they occasionally read or write humorous comments [ex10]. [ex10] Sometimes, more or less, jokes are told, (…) and then it's definitely entertaining. (m, 26, occ.) For most participants, humorous comments provided some entertainment value. Only a few interviewees perceived these comments as an opportunity to join in the discussion with a similar comment. Most participants, however, did not respond to these comments or discuss them. Formal descriptors Many participants reported that reading other users' comments sometimes is a time-consuming activity. Both the sheer number of comments to some news items and the length of single comments exhausted their cognitive resources. With regard to length, the participants concluded that the readers of their comments also have limited resources and prefer reading and responding to short comments [ex11a]. [ex11a] I write short (…) because I feel that people are hacked off when there is something lengthier in the discussion. (m, 23, reg.) On the other hand, some users argued that it requires time and space to contribute a well-argued comment to the discussion. Furthermore, some participants perceived that long comments usually include more “contact points” for an interactive discussion. While the length of a comment was judged differently, all of the participants agreed that the position of a comment determines to a high degree whether it would receive a response [ex11b]. [ex11b] I once wrote a late comment (…) and that's why I didn't get much feedback. (m, 20, reg.) Consistently, the participants perceived that response comments primarily address comments that appear early in the discussion. Discourse architecture When talking about their specific perceptions of online news discussions, the participants described their own platform-specific discussion behavior (RQ2). Concerning the sociability of online communication services (Preece, 2001), many users assumed that their commenting behavior was guided by platform-specific degrees of anonymity, behavioral rules, or codes of conduct. Most remarkably, some interviewees claimed to respond to aggressive and controversial comments only on platforms that grant anonymity [ex12a]. [ex12a] I think, if I wanted to comment on something controversial, then I would do that rather on the websites, where the profile is not visible. (f, 21, occ.) Differences in the degree of usability between the communication services often concerned registration rules. These rules were perceived as an inhibitor of responding to other comments, mainly because of disproportional time and effort requirements. Furthermore, the participants perceived that their response behavior was affected by the way comments are sorted on the platforms because this order determines the comments they are confronted with at first [ex12b]. [ex12b] On Facebook, you usually see the latest two comments of a discussion (…), and, personally, I mostly (…) refer to the most recent ten comments. (m, 26, occ.) In summary, the participants' responses indicate that the sociability of a communication service primarily affects whether or not they reply to comments with specific discussion factors, while the usability primarily affects their general response behavior and the overall number of comments they perceive as relevant (cf. also Preece, 2001). Discussion On the basis of the perceptions and statements of 25 users who comment on news items, Study 1 generated a catalog of discussion factors. Many of these factors are consistent with traditional news factors, although some of their specific characteristics resemble “quality factors” from online deliberation research (e.g., questions, expertise) (Papacharissi, 2004; Ruiz et al., 2011) and “usefulness factors” from research on user-generated product reviews (e.g., additional factual information) (Walther, & Jang, 2012; Willemsen, Neijens, Bronner, & de Ridder, 2011). Our findings underpin the assumption that users consider factors on multiple message dimensions when thinking about whether or not to reply to a previous comment. One remarkable finding is that the users we interviewed differentiated between aggressive behavior and controversial statements. The discussion factor aggression primarily refers to hostile comments that solely aim at inflicting harm to the discussants and thus resembles the phenomenon of “flaming” (e.g., Alonzo & Aiken, 2004) or what Papacharissi (2004) calls “impoliteness.” In contrast, the factor controversy includes topic-related statements that are very likely to stimulate disagreement, some of them resembling what scholars have described as “incivility” or “trolling” (e.g. Herring et al., 2002; Papacharissi, 2004). Regarding the effects of the discussion factors, our results can again be linked to news value theory: Controversy, for example, challenged the normative beliefs of some of the participants and thereby suggested the social significance of “balance” in the public sphere by voicing another (opposing) viewpoint (cf. also Price, Nir, & Cappella, 2006; Shoemaker & Cohen, 2006). Other discussion factors, such as uncertainty, primarily seem to activate a reader's prior (factual) knowledge and thereby increase the probability of stimulating responses. Finally, the effect of factors such as humor can be linked to an individual's personal well-being because it creates a sense of hedonic enjoyment. Yet, a limitation of Study 1 is that it could only extract perceived discussion factors that were salient to the participants. While this approach can be interpreted as an alternative to normative and deductive analyses, additional factors might guide user behavior. Future research might want to complement our approach by asking users to take the researcher through their preferred platform and to talk aloud about the characteristics of news items and user comments that increase their need to comment on them. With this limitation in mind, Study 1 has provided some insights into why some comments receive response comments from other users (i.e., create interactivity). Regarding the direction of influence of the extracted discussion factors and the effects of different discourse architectures on the structure of the discussions, it has generated ambivalent results. Therefore, we conducted a second study with the aim of quantifying the effects of the discussion factors identified in Study 1. Study 2 The aim of the content analysis of Study 2 was to analyze which of the discussion factors identified in Study 1 determine whether a user comment receives response comments (i.e., feedback) from other users (RQ3). Additionally, we aimed at providing preliminary insights into the interplatform validity of the discussion factors. In particular, the results from our qualitative study suggested that some users respond to aggressive and controversial comments only on platforms that grant anonymity. Likewise, previous research has shown that anonymity can be linked to a more unconcerned or uninhibited opinion expression behavior (e.g., Suler, 2004; Yun & Park, 2011). On the basis of these findings, we hypothesized that aggressive and controversial comments would be more likely to receive response comments on platforms that allow anonymous or pseudonymous commenting than on platforms that require real-name registration (H1). Furthermore, the respondents of our qualitative study had emphasized that the visibility of comments affects their response behavior: Comments appearing late in a discussion were likely to be overlooked on platforms where comments are sorted chronologically. In keeping with this, for traditional news items too, formal message components, such as its position, have also repeatedly been found to guide readers' selection behavior and the amount of importance they assign to them (cf. Eilders, 2006). Based on these findings, we hypothesized that comments that appear late in a discussion would be more likely to receive responses on platforms where comments are sorted in reverse chronological order than on platforms where comments are sorted in chronological order (H2). To answer the research question and determine the validity of the two hypotheses, we conducted a content analysis of user comments on two German news websites and on their respective Facebook pages. Sample The websites analyzed were Spiegel Online (www.spiegel.de), Bild.de (www.bild.de), and their respective Facebook pages. Both websites were chosen (a) because of the large number of users who had “liked” their Facebook pages and (b) because of their high popularity in Germany. At the time of data collection, the two websites were the most visited German news-only websites (Alexa, 2013). Furthermore, at the time of data collection, the discourse architectures of Bild.de and Spiegel Online were similar on the web platforms (users could register with a pseudonym, comments were premoderated and sorted chronologically, with 10 comments appearing per page), but they differed from their Facebook pages (users were asked to disclose their real name, comments were postmoderated and sorted in reverse chronological order, with the latest two comments displayed). However, both platforms provided only one response level (i.e., users had to cite other comments manually in order to respond to them). Altogether, 1,580 user comments on 18 political news articles were coded on Spiegel Online, Bild.de, and on their respective Facebook pages. The procedure was as follows: Starting from the Facebook pages, a sample of news articles was randomly selected during 2 weeks in late October 2011. The news articles on the Facebook pages referred to the respective articles on the news websites. Hence, for each Facebook news article in our sample, the respective article on the news website was included. Website and Facebook news articles that were not published in the “politics” section on the news websites were excluded from the sample. We decided to analyze only political news items because our interviews in Study 1 and previous research have shown that users comment on political issues most frequently (e.g., Boczkowski & Mitchelstein, 2012; Ruiz et al., 2011). However, a news item was included in the final sample only when the respective user discussions included at least 20 comments each on the Facebook page and on the respective news website. We decided in favor of a minimum of 20 comments per discussion because previous research has shown that between 20 and 50% of user comments receive response comments (Ruiz et al., 2011; Singer, 2009). In a sample of at least 20 comments, we expected to find sufficient user-to-user-interactions to answer our research question and validate the hypotheses. In order to capture a greater variety of discussion topics, the number of comments analyzed per news item had to be restricted to 100. Therefore, discussions with more comments were not analyzed completely. Each discussion was saved in a text file to guarantee that there were no changes during the coding process. Coding The coding scheme for the content analysis included 10 discussion factors (cf. Table 1), four formal descriptors (position, length, medium, and platform for user comments), and several other categories such as a comment's publishing time and the number of response comments it had received. The category descriptions were adopted from the participants' perceptions of the discussion factors in Study 1 (cf. Table 1). The coding was based on a procedure suggested by Papacharissi (2004): For most discussion factors, a multiterm index was used to code for the presence or absence of the respective discussion factor. For example, if there was at least one controversial statement in a user comment (i.e., unfounded demands, provocative questions, exaggerations, or stereotypes), then this was considered sufficient for it to be coded as controversy (cf. coding scheme in Table 1). The multiterm indices were built upon the statements of the participants of Study 1. For analytical purposes, the discussion factors were coded dichotomously. In other words, we did not differentiate between comments that included one controversial statement and comments that included multiple controversial statements. While future studies might apply more precise measures, the current study was exploratory in its nature, and an additive effect of the single indicators of discussion factors could not be assumed definitely. For example, a comment that insults other users and simultaneously “shouts” by using all caps might not be perceived as more offensive than a comment that includes only one indicator of aggression (cf. Papacharissi, 2004). Table 1 Coding Scheme, Intercoder Agreements, and Frequency of the Variables Used in the Analysisa Discussion Factor a . Coding Scheme . Frequency . PA . α . Aggression (1) Using derogatory language and cuss words (insults) 10% .95 .76 (2) Accusing others of lacking competency in doing something (accusation of incompetence) (3) Using all-caps (“shouting”) Facticity (1) Claiming to be an expert or insider (expertise) 4% 1 —c (2) Including proved (i.e., linked) factual knowledge (factual knowledge) Uncertainty Including questions that visibly aim at closing the author's information or knowledge gaps 9% .98 .80 Controversy (1) Reflecting the author's generalized beliefs about the members of specific groups (stereotypes) 34% .89 .78 (2) Including unfounded demands (3) Asking provocative—often rhetorical—questions (e.g., “Who cares for this irrelevant news?”). Expected responses to these questions are more in the form of agreement or denial, and less information transfer (4) Extensively using superlatives or making things seem bigger, more important, or worse than they are (exaggerations) Unexpectednessd Providing a different perspective on the issue under discussion that was neither mentioned in the news article nor in the previous comments (e.g., offering alternative interpretations, re-contextualizing an issue, etc.) 23% .86 .63 Personalizatione Directly addressing one or more participants of the discussion 24% .99 .90 Simplification (1) Assuming the consequences of an event to stem from only one definite cause 6% .94 .71 (2) Stating that an issue is less complex than described in the news article Humor Making jokes or using smileys so as to clarify that the statement is meant to be a joke 9% 1 1 Comprehensibility (1) Substituting a word with an expression from another subject area. The expression has a meaning that is assigned to the original word (metaphor) 40% .86 .72 (2) Meaning the opposite of what is said (irony) Negativity Writing in an exclusively negative tone 40% .87 .74 Length Number of words of the comment without counting direct quotations. Comments were divided into four categories: very short (max. 10 words), short (max. 20 words), medium (max. 47 words), long (more than 47 words). — .94 .99 Position Position of the comment in the discussion in chronological order. The comments were recoded into three categories: beginning, middle, end. — 1 1 Discussion Factor a . Coding Scheme . Frequency . PA . α . Aggression (1) Using derogatory language and cuss words (insults) 10% .95 .76 (2) Accusing others of lacking competency in doing something (accusation of incompetence) (3) Using all-caps (“shouting”) Facticity (1) Claiming to be an expert or insider (expertise) 4% 1 —c (2) Including proved (i.e., linked) factual knowledge (factual knowledge) Uncertainty Including questions that visibly aim at closing the author's information or knowledge gaps 9% .98 .80 Controversy (1) Reflecting the author's generalized beliefs about the members of specific groups (stereotypes) 34% .89 .78 (2) Including unfounded demands (3) Asking provocative—often rhetorical—questions (e.g., “Who cares for this irrelevant news?”). Expected responses to these questions are more in the form of agreement or denial, and less information transfer (4) Extensively using superlatives or making things seem bigger, more important, or worse than they are (exaggerations) Unexpectednessd Providing a different perspective on the issue under discussion that was neither mentioned in the news article nor in the previous comments (e.g., offering alternative interpretations, re-contextualizing an issue, etc.) 23% .86 .63 Personalizatione Directly addressing one or more participants of the discussion 24% .99 .90 Simplification (1) Assuming the consequences of an event to stem from only one definite cause 6% .94 .71 (2) Stating that an issue is less complex than described in the news article Humor Making jokes or using smileys so as to clarify that the statement is meant to be a joke 9% 1 1 Comprehensibility (1) Substituting a word with an expression from another subject area. The expression has a meaning that is assigned to the original word (metaphor) 40% .86 .72 (2) Meaning the opposite of what is said (irony) Negativity Writing in an exclusively negative tone 40% .87 .74 Length Number of words of the comment without counting direct quotations. Comments were divided into four categories: very short (max. 10 words), short (max. 20 words), medium (max. 47 words), long (more than 47 words). — .94 .99 Position Position of the comment in the discussion in chronological order. The comments were recoded into three categories: beginning, middle, end. — 1 1 a Due to a lack of space, the full coding scheme cannot be presented here in detail. However, it can be provided by the authors on request. b All discussion factors were coded dichotomously, unless otherwise noted. c Facticity showed no variation in the results for intercoder agreement. d The indicator ‘topic drifts’ was defined too imprecisely and thus had to be dropped. e Due to an error in the coding scheme, only the addresses made by other users could be coded. The coding of whether a user was concerned personally by a news item was not included in the variable. Open in new tab Table 1 Coding Scheme, Intercoder Agreements, and Frequency of the Variables Used in the Analysisa Discussion Factor a . Coding Scheme . Frequency . PA . α . Aggression (1) Using derogatory language and cuss words (insults) 10% .95 .76 (2) Accusing others of lacking competency in doing something (accusation of incompetence) (3) Using all-caps (“shouting”) Facticity (1) Claiming to be an expert or insider (expertise) 4% 1 —c (2) Including proved (i.e., linked) factual knowledge (factual knowledge) Uncertainty Including questions that visibly aim at closing the author's information or knowledge gaps 9% .98 .80 Controversy (1) Reflecting the author's generalized beliefs about the members of specific groups (stereotypes) 34% .89 .78 (2) Including unfounded demands (3) Asking provocative—often rhetorical—questions (e.g., “Who cares for this irrelevant news?”). Expected responses to these questions are more in the form of agreement or denial, and less information transfer (4) Extensively using superlatives or making things seem bigger, more important, or worse than they are (exaggerations) Unexpectednessd Providing a different perspective on the issue under discussion that was neither mentioned in the news article nor in the previous comments (e.g., offering alternative interpretations, re-contextualizing an issue, etc.) 23% .86 .63 Personalizatione Directly addressing one or more participants of the discussion 24% .99 .90 Simplification (1) Assuming the consequences of an event to stem from only one definite cause 6% .94 .71 (2) Stating that an issue is less complex than described in the news article Humor Making jokes or using smileys so as to clarify that the statement is meant to be a joke 9% 1 1 Comprehensibility (1) Substituting a word with an expression from another subject area. The expression has a meaning that is assigned to the original word (metaphor) 40% .86 .72 (2) Meaning the opposite of what is said (irony) Negativity Writing in an exclusively negative tone 40% .87 .74 Length Number of words of the comment without counting direct quotations. Comments were divided into four categories: very short (max. 10 words), short (max. 20 words), medium (max. 47 words), long (more than 47 words). — .94 .99 Position Position of the comment in the discussion in chronological order. The comments were recoded into three categories: beginning, middle, end. — 1 1 Discussion Factor a . Coding Scheme . Frequency . PA . α . Aggression (1) Using derogatory language and cuss words (insults) 10% .95 .76 (2) Accusing others of lacking competency in doing something (accusation of incompetence) (3) Using all-caps (“shouting”) Facticity (1) Claiming to be an expert or insider (expertise) 4% 1 —c (2) Including proved (i.e., linked) factual knowledge (factual knowledge) Uncertainty Including questions that visibly aim at closing the author's information or knowledge gaps 9% .98 .80 Controversy (1) Reflecting the author's generalized beliefs about the members of specific groups (stereotypes) 34% .89 .78 (2) Including unfounded demands (3) Asking provocative—often rhetorical—questions (e.g., “Who cares for this irrelevant news?”). Expected responses to these questions are more in the form of agreement or denial, and less information transfer (4) Extensively using superlatives or making things seem bigger, more important, or worse than they are (exaggerations) Unexpectednessd Providing a different perspective on the issue under discussion that was neither mentioned in the news article nor in the previous comments (e.g., offering alternative interpretations, re-contextualizing an issue, etc.) 23% .86 .63 Personalizatione Directly addressing one or more participants of the discussion 24% .99 .90 Simplification (1) Assuming the consequences of an event to stem from only one definite cause 6% .94 .71 (2) Stating that an issue is less complex than described in the news article Humor Making jokes or using smileys so as to clarify that the statement is meant to be a joke 9% 1 1 Comprehensibility (1) Substituting a word with an expression from another subject area. The expression has a meaning that is assigned to the original word (metaphor) 40% .86 .72 (2) Meaning the opposite of what is said (irony) Negativity Writing in an exclusively negative tone 40% .87 .74 Length Number of words of the comment without counting direct quotations. Comments were divided into four categories: very short (max. 10 words), short (max. 20 words), medium (max. 47 words), long (more than 47 words). — .94 .99 Position Position of the comment in the discussion in chronological order. The comments were recoded into three categories: beginning, middle, end. — 1 1 a Due to a lack of space, the full coding scheme cannot be presented here in detail. However, it can be provided by the authors on request. b All discussion factors were coded dichotomously, unless otherwise noted. c Facticity showed no variation in the results for intercoder agreement. d The indicator ‘topic drifts’ was defined too imprecisely and thus had to be dropped. e Due to an error in the coding scheme, only the addresses made by other users could be coded. The coding of whether a user was concerned personally by a news item was not included in the variable. Open in new tab Intercoder reliability between the four coders was tested for all categories during January 2014. Owing to the exploratory nature of the study, a Krippendorff's α value of .67 was considered as the minimum value acceptable to draw tentative conclusions (Krippendorff, 2004; Lombard, Snyder-Duch, & Bracken, 2002). Table 1 describes the coding scheme and reports both the Krippendorff's α and percent agreement values (Holsti's method), and thereby includes multiple intercoder indices as per the suggestion of previous reports (e.g., Lombard et al., 2002). Table 1 also reports the frequency of the variables in the final dataset. All variables except unexpectedness (α = .63) had a Krippendorff's α greater than .70 (cf. Table 1). Although earlier studies on news factors reported similar difficulties when coding unexpectedness (Weber, 2013) and although the percent agreement-score of the factor was sufficient (PA = .86), doubts about the validity of the category seem legitimate and the following results should therefore be interpreted with caution. Results RQ3 was regarding the influence of the discussion factors on whether or not a comment receives response comments (i.e., feedback). To answer this question, we counted the comments that visibly responded to one of the 1,580 comments analyzed and retained this information in the variable feedback. Overall, 268 initial comments received 407 response comments: 76% of the 268 feedback-stimulating comments received one response, 16% received two responses, and only 8% of the feedback-stimulating comments received more than two responses. Owing to this comparatively low number of comments that received more than one response comment, the variable feedback was recoded dichotomously (0 = no feedback received, 1 = feedback received). We then computed a binary logistic regression model. The variable feedback was entered as the dependent variable. The discussion factors from Study 1 and the formal comment descriptors (length and position) were entered as predictor variables. The platform (WWW vs. Facebook) and the medium variables (Bild.de vs. Spiegel Online) were entered as control variables. Additionally, previous research has shown that the characteristics of news stories can influence the interactivity of news discussions (Weber, 2013). To control whether these characteristics also influence the effects of the discussion factors, we entered a very basic indicator of news characteristics (i.e., the topics of the news items) into our regression model. The results are shown in the first column of Table 2 (model I). Table 2 Logistic Regression Analysis of the Effect of Discussion Factors on Comment Feedback Probability Item . Model I (RQ3) Main Effects (effect of discussion factors on comment feedback probability) . Model II (H1 & H2) Hypothesized INteraction effects (factors ×  platform) . b-value . Odds . b-value . Odds . Constant .99* 2.71  .28 1.32  Block 1: Discussion factors Aggression .44† 1.55   1.37* 3.92  Controversy .56** 1.74 −.17 .84 Facticity −.04    .96 Unexpectedness .49** 1.63 Negativity −.36*    .70 Personalization .47*  1.60 Simplification .08   1.08 Comprehensibility .40*  1.36 Uncertainty  .78** 2.18 Humor −.35    .71 Block 2: Formal descriptors Length: Long (Ref. cat.) — — Length: Medium −.15   .86 Length: Short −.22   .80 Length: Very short −.91**  .41 Position: End (Ref. cat.) — — — — Position: Middle .37  1.45  .40 1.49 Position: Beginning   .93*** 2.52  .26 1.29 Block 3: Medium and platform variables News medium (1 = Bild.de) −1.86*** .16  Platform (1 = Facebook) −.38*  .68  Block 4: News story variables T1 U.S. arms policy (Ref. cat.) — — T2 Euro rescue fund −.17   .85  T3 Downsizing of the Bundeswehr −.18   .84  T4 Announcement of tax cut in Germany −1.01   .36  T5 Minimum wages in Germany  1.27** 3.57  T6 WikiLeaks shutdown .56 1.75  R2 (Cox & Snell) .18 .19 R2 (Nagelkerke) .30 .33 χ2 (Model) 310.46*** 253.25*** Item . Model I (RQ3) Main Effects (effect of discussion factors on comment feedback probability) . Model II (H1 & H2) Hypothesized INteraction effects (factors ×  platform) . b-value . Odds . b-value . Odds . Constant .99* 2.71  .28 1.32  Block 1: Discussion factors Aggression .44† 1.55   1.37* 3.92  Controversy .56** 1.74 −.17 .84 Facticity −.04    .96 Unexpectedness .49** 1.63 Negativity −.36*    .70 Personalization .47*  1.60 Simplification .08   1.08 Comprehensibility .40*  1.36 Uncertainty  .78** 2.18 Humor −.35    .71 Block 2: Formal descriptors Length: Long (Ref. cat.) — — Length: Medium −.15   .86 Length: Short −.22   .80 Length: Very short −.91**  .41 Position: End (Ref. cat.) — — — — Position: Middle .37  1.45  .40 1.49 Position: Beginning   .93*** 2.52  .26 1.29 Block 3: Medium and platform variables News medium (1 = Bild.de) −1.86*** .16  Platform (1 = Facebook) −.38*  .68  Block 4: News story variables T1 U.S. arms policy (Ref. cat.) — — T2 Euro rescue fund −.17   .85  T3 Downsizing of the Bundeswehr −.18   .84  T4 Announcement of tax cut in Germany −1.01   .36  T5 Minimum wages in Germany  1.27** 3.57  T6 WikiLeaks shutdown .56 1.75  R2 (Cox & Snell) .18 .19 R2 (Nagelkerke) .30 .33 χ2 (Model) 310.46*** 253.25*** Note: n = 1,580; 0 = no feedback received, 1 = feedback received; *** p < .001; ** p < .01; * p < .05; † p < .1. Open in new tab Table 2 Logistic Regression Analysis of the Effect of Discussion Factors on Comment Feedback Probability Item . Model I (RQ3) Main Effects (effect of discussion factors on comment feedback probability) . Model II (H1 & H2) Hypothesized INteraction effects (factors ×  platform) . b-value . Odds . b-value . Odds . Constant .99* 2.71  .28 1.32  Block 1: Discussion factors Aggression .44† 1.55   1.37* 3.92  Controversy .56** 1.74 −.17 .84 Facticity −.04    .96 Unexpectedness .49** 1.63 Negativity −.36*    .70 Personalization .47*  1.60 Simplification .08   1.08 Comprehensibility .40*  1.36 Uncertainty  .78** 2.18 Humor −.35    .71 Block 2: Formal descriptors Length: Long (Ref. cat.) — — Length: Medium −.15   .86 Length: Short −.22   .80 Length: Very short −.91**  .41 Position: End (Ref. cat.) — — — — Position: Middle .37  1.45  .40 1.49 Position: Beginning   .93*** 2.52  .26 1.29 Block 3: Medium and platform variables News medium (1 = Bild.de) −1.86*** .16  Platform (1 = Facebook) −.38*  .68  Block 4: News story variables T1 U.S. arms policy (Ref. cat.) — — T2 Euro rescue fund −.17   .85  T3 Downsizing of the Bundeswehr −.18   .84  T4 Announcement of tax cut in Germany −1.01   .36  T5 Minimum wages in Germany  1.27** 3.57  T6 WikiLeaks shutdown .56 1.75  R2 (Cox & Snell) .18 .19 R2 (Nagelkerke) .30 .33 χ2 (Model) 310.46*** 253.25*** Item . Model I (RQ3) Main Effects (effect of discussion factors on comment feedback probability) . Model II (H1 & H2) Hypothesized INteraction effects (factors ×  platform) . b-value . Odds . b-value . Odds . Constant .99* 2.71  .28 1.32  Block 1: Discussion factors Aggression .44† 1.55   1.37* 3.92  Controversy .56** 1.74 −.17 .84 Facticity −.04    .96 Unexpectedness .49** 1.63 Negativity −.36*    .70 Personalization .47*  1.60 Simplification .08   1.08 Comprehensibility .40*  1.36 Uncertainty  .78** 2.18 Humor −.35    .71 Block 2: Formal descriptors Length: Long (Ref. cat.) — — Length: Medium −.15   .86 Length: Short −.22   .80 Length: Very short −.91**  .41 Position: End (Ref. cat.) — — — — Position: Middle .37  1.45  .40 1.49 Position: Beginning   .93*** 2.52  .26 1.29 Block 3: Medium and platform variables News medium (1 = Bild.de) −1.86*** .16  Platform (1 = Facebook) −.38*  .68  Block 4: News story variables T1 U.S. arms policy (Ref. cat.) — — T2 Euro rescue fund −.17   .85  T3 Downsizing of the Bundeswehr −.18   .84  T4 Announcement of tax cut in Germany −1.01   .36  T5 Minimum wages in Germany  1.27** 3.57  T6 WikiLeaks shutdown .56 1.75  R2 (Cox & Snell) .18 .19 R2 (Nagelkerke) .30 .33 χ2 (Model) 310.46*** 253.25*** Note: n = 1,580; 0 = no feedback received, 1 = feedback received; *** p < .001; ** p < .01; * p < .05; † p < .1. Open in new tab According to the data, several discussion factors significantly influenced the probability of feedback on a comment: Comments were more likely to receive response comments when they included controversial statements (controversy), when they addressed other users in order to emphasize why the comment is relevant to them (personalization), when they asked questions to close gaps in their own knowledge (uncertainty), and when they provided a different perspective on the issue under discussion (unexpectedness). Moreover, user comments that did not include irony or metaphors (comprehensibility), and that were not exclusively negative in their tone (negativity) were more likely to receive responses. The trend indicated that offensive comments (aggression) were more likely to receive response comments. With regard to the length of comments, only very short comments (up to 10 words) were less likely to receive response comments. Regarding the position of comments, comments that were published in the first third of a discussion were significantly more likely to receive feedback than comments posted in the last third (end). Additionally, comments on Spiegel Online were more likely to receive a response than comments on Bild.de. With regard to the discourse architectures, comments posted on the website platforms of the news media were more likely to receive response comments than comments posted on Facebook. Finally, comments posted as a response to the topic “minimum wages in Germany” were more likely to receive response comments. Platform differences To test the validity of H1 and H2, we additionally entered the interactions between the factors controversy, aggression, and position and the platform (Facebook or websites) into the binary logistic regression model.6 The interactions were created by multiplying the item platform with one of the four factors from the same column. The results are shown in the second column of Table 2 (model II). The interaction effect was significant only for the discussion factor aggression. However, contrary to H1, aggressive comments were more likely to receive feedback if posted on Facebook rather than on the website. The effect of controversy was not influenced by the discourse architectures at all. Likewise, comments located in the middle and end of discussions were not more likely to stimulate feedback on Facebook. Therefore, both H1 and H2 were rejected. Discussion The current study used the rationale of news value theory to argue that user comments contain so-called discussion factors which—via various cognitive and affective routes—increase the perceived relevance of the comments and therefore increase the likelihood of a later user responding to such comments. Qualitative interviews revealed that users indeed recognize factors that increase their motivation to respond to other users' comments. These factors were classified with the help of the traditional news factor framework. Using quantitative content analysis, we then explored the factors that were successful in stimulating feedback. The results of this multistep approach implied that the authors of user comments can trigger response comments by including controversy, unexpectedness, personalization, and uncertainty in their postings and by avoiding incomprehensibility and negativity. Length, position, the news medium itself, and the news story topic further affected the probability of whether a comment received feedback. In the following paragraphs, we will try to provide some arguments as to why the discussion factors that were confirmed in our quantitative content analysis were effective in stimulating user-to-user-interactions. We found that comments including controversy were about 1.7 times as likely as comments not including this factor to stimulate feedback. This result parallels findings from audience-oriented news value research that controversial news items are more likely to be selected and commented upon by news media users (Eilders, 2006; Boczkowski & Mitchelstein, 2012). Controversy thus appears as a rather general trigger of cognitive and behavioral responses in news selection and public news discussion. An explanation for this might be that comments including provocative questions, stereotypes, unfounded demands, or exaggerations are likely to deviate from a subsequent commenter's individual or internalized opinion, norms, and/or values. Deviant messages then might not only increase the amount of attention paid to the message (cf. Shoemaker, 1996) but also the sense of disagreement. A sense of disagreement, in turn, appears as a strong trigger to write response comments (Diakopoulos & Naaman, 2011; Singer, 2009; Springer & Pfaffinger, 2012): For example, Boczkowski & Mitchelstein (2012) described commenting users as “monitorial citizens” (p. 15) who control their environment and, when facing disagreement, feel the need to “broadcast” their opinion publicly. Conceptualizing controversy as a general indicator of discussion value is somewhat supported by the nonsignificant interaction effect between this discussion factor and different discourse architectures—in fact, controversial comments stimulated response comments both on platforms that grant relative anonymity (WWW) and on platforms that require real-name registration (Facebook). This result is in line with previous research that found users' posting behavior not to be influenced by the sense of anonymity: “people posted a message when they felt that it was necessary, even though doing so required them to reveal their personal identity to the forum administrators” (Yun & Park, 2011, p. 216). To some extent, the second study supported the argument that users differentiate between controversial and aggressive comments. While controversy clearly stimulated response comments, users responded to blunt and offensive personal offenses (aggression) primarily on Facebook. On this platform, aggressive comments were nearly four times as likely as aggressive comments posted on the websites to stimulate feedback. This finding contradicts H1 and the perceptions of the users interviewed in Study 1. However, the effect might be explained by using the social identity model of depersonalization effects. This model suggests that individual users tend to orient their perceptions and behaviors on the basis of perceived similarities with the characteristics of a group as a whole rather than on the basis of their individual similarities and differences with specific users participating in the communication (e.g., Postmes, Spears, & Lea, 1998). When only a minority of the participants in anonymous news discussions on mass media websites writes aggressive comments (only 1 in 10 comments in the current sample) or responds to them, it might be rational for a single user to ignore these comments, even though such behavior might contradict individual norms. On Facebook, however, discussants typically offer a higher number of authentic and visual cues to their identity (e.g., Trepte & Reinecke, 2011). Here, the condition of visual anonymity is not granted and readers of provocative comments can discern the provocateur's similarities and differences based on their self-perception. As a consequence, they might be more likely to respond to aggressive comments that violate their individual normative beliefs.7 Nevertheless, future research should investigate this effect by adding more news media and comments to their samples. The second study further confirmed that users prefer responding to comments that were not written in an exclusively negative tone. Selecting comments for discussion obviously works differently than news selection, as predicted by news value theory (Galtung & Ruge, 1965). Users seem to perceive the authors of more neutral or positively written comments as more agreeable discussion partners than the authors of exclusively negative comments. While research on deliberation processes emphasizes the importance of topic-centered discussions (e.g., Freelon, 2010; Stromer-Galley & Martinson, 2009), the current study found that comments that provide a different perspective toward the issue under discussion (e.g., an alternative interpretation) were more likely to stimulate feedback than comments that strictly stayed on track. It can be argued that such unexpected comments were perceived as new structuring topics (e.g., Stromer-Galley, 2007) that supplied the conversation with additional aspects worth discussing. Such additional information can increase a subsequent user's involvement and tie in with pre-existing knowledge that he/she can contribute to the discussion. However, the finding may also be interpreted as an indicator of the self-regulating characteristic of online discussions: Unexpected comments might stimulate some kind of “metatalk” (Stromer-Galley & Martinson, 2009, p. 200) intended to rebuke previous users for not strictly keeping to the original topic. Although we cannot distinguish between these interpretations, unexpectedness primarily seems to stimulate responses when respondents can link the unexpected information with pre-existing cognitive schemes. In this regard, the factor corroborates the findings of news value theory, which argued that unexpectedness can be effective only if it occurs within an individual's relevant set of topics perceived as important and meaningful (Galtung & Ruge, 1965). Personalizing comments by explicitly addressing other users and creating “controllable” uncertainty by asking questions increased the probability of receiving feedback. These effects might be interpreted from a sociocognitive news value perspective: Early studies argued that personalized news items allow individuals to easily identify with the events and circumstances reported (Galtung & Ruge, 1965). Similarly, addressing other users might help these users to recognize the personal relevance of the message and increase their perceived self-efficacy to respond (cf. Study 1). Regarding the discussion factor uncertainty, asking questions implies an expectation of the sender(s) to get a response and—given sufficient knowledge of the receiver(s)—they commit them to some extent to provide this response (Schank, 1977; Stromer-Galley, 2007). Asking questions might also motivate so-called “answer persons” to join the discussion. These users are well recognized across a variety of online discussion spaces and they have both the knowledge and the altruism to answer questions (Turner, Smith, Fisher, & Welser, 2005). With regard to formal descriptors of user comments, only very short comments were less likely to receive feedback than other comments. This result is counterintuitive at first sight because most participants in Study 1 asserted that they prefer writing and reading short comments. However, when comments are too short they cannot contain much useful information, that is, new aspects that keep the discussion going. Furthermore, long comments probably include additional factors not covered by this analysis, for example, the quality of the arguments used. The position of a comment had a noticeable impact on whether it was intensively responded to or not; comments posted at the beginning or the middle of a discussion were significantly more likely to stimulate response comments. While it is plausible to assume that earlier comments are better able to frame (Entman, 1993) the discussion than later ones, this result might partly be a methodological artifact. Comments at the end of the discussion cannot—by definition—receive as many response comments as those which started the discussion. However, we only analyzed the first 100 comments, and in many cases there were more comments that appeared later. As a result, the position “end” does not necessarily describe the final stage of a discussion. Comments on news items of rather serious news media (Spiegel Online) were more likely to receive response comments than comments on news items of boulevard media (Bild.de). Numerous interpretations seem adequate to explain this difference, including differences in the behavior of different news audiences and different restrictiveness of the journalistic premoderation rules. Regarding the influence of the news story topic, our results seem to confirm that users preferably engage in interactive discussions when a topic—such as the introduction of a nationwide minimum wage in Germany—concerns their own nation and has a large impact on a specified group (Weber, 2013). However, the results also imply that the discussion factors help explain interactivity in news discussions beyond the topic of a news story. Finally, the second study provided limited support for the hypothesis that different technological features of the discourse architecture alter the discussion processes. First, comments on the web platforms were more likely to stimulate responses than comments on Facebook. It might therefore be possible that users perceive Facebook's comment function not so much as a utility to discuss issues of public relevance interactively but rather to add their two cents to the conversation. Additionally, journalists on Facebook often explicitly invite users to voice their opinion (e.g., “What do you think about it?”). This aspect of sociability might cause users to primarily respond to the specific journalistic question instead of responding to the comments of other users. Second, regarding H2, different comment sequencing techniques did not influence the likelihood of late or early comments receiving responses. These results somehow contradict the findings from our qualitative research and the more technological-deterministic arguments from previous research (e.g., Jones & Rafaeli, 2000; Wright & Street, 2007). Nevertheless, we only examined a very limited variety of different discourse architectures and, more specifically, we focused on one specific aspect of different technical implementations of the comment function (chronological vs. reverse chronological order). Future research should investigate whether other sequencing techniques (e.g., displaying readers' or editors' choices of comments) or other discussion features (e.g., the availability of a “reply-button”) affect the discussion structure. From a more general perspective, the current study also advances research on interactivity: By analyzing the characteristics of “initial” user messages that spur reciprocity between users in online news discussions, it complements earlier research on interactivity-as-process (e.g., Stromer-Galley, 2004) which analyzed the formal and content characteristics of messages that respond to “initial” user messages (e.g., Rafaeli & Sudweeks, 1997). Further, our consideration of the different discourse architectures shows that research on interactivity-as-process should not disregard the moderating influence of the interactivity of the technological features, that is, the interactivity-as-product variables, respectively (Stromer-Galley, 2004). To some extent, these variables can affect the prevalence and the nature of user-to-user interactions. These findings should encourage researchers to continue investigating when and why “the degree or features of medium interactivity might affect outcome variables of human interaction” (Stromer-Galley, 2004, p. 393; Wright & Street, 2007). Ultimately, the current study illustrates that in order to establish a theoretical bridge between technological features and processes, the perception of interactivity of the users has to be taken into account (e.g., Quiring, 2009). Limitations The current study can be seen as a first exploratory step into the concept of discussion factors. As such, it contains several limitations that should be addressed in future research: First of all, we cannot be sure that the list of factors we derived is complete. Additional factors guiding individual user behavior might not have been captured with the method of qualitative interviews. Second, we considered only a very simple indicator (topic) of the original news characteristics to control the influence of the news items on the discussion structures. Earlier research has shown that different news items can stimulate different degrees of interactivity (Weber, 2013). Future studies hence should analyze the joint influence of news factors in news items and discussion factors in user comments in more detail. Third, in order to test the robustness of the effects reported, future studies should analyze a bigger (and maybe more international) sample. Fourth, only 268 of the 1,580 comments in our sample received at least one response comment. Voicing no response or few responses thus are the main reactions to comments of other users. However, our coding was restricted to the first 100 comments, and comments might have received more responses later on. But even when a comment receives only one response comment, this might indicate to a subsequent reader that this particular comment is more worthy to be thought about than others. Fifth, the coding procedure of the discussion factors could be improved. Owing to low reliability, two indicators of the discussion factors from Study 1 were not analyzed in Study 2 (cf. Table 1). Most importantly, future studies should define precisely what they consider as a “topic drift” (unexpectedness). Additionally, instead of using multiterm indices, future studies might code each indicator separately or establish a “factor intensity index.” Sixth, we did not comprehensively analyze the origins of the discussion factors. We assume that these factors are stimulated by a combination of specific news factors in news items and specific commenting practices. However, it remains an open question to what extent news factors and commenting practices inform discussion factors. Seventh, our analysis does not allow inferring whether the discussion factors stimulated quality debates or just harsh rebukes. Future research should include more specific aspects of the response comments (e.g., Rafaeli & Sudweeks, 1997), for example, whether they criticize or support the user they respond to. Finally, our measure of interactivity was very simple. Whether a comment receives a (semantically linkable) response comment is a very basic indicator and can only shed first light on the complexity of real-world interactions. Yet, it was the aim of the current study to adhere to the logic of news value theory, which is to explain a discrete cognitive or behavioral response to a news item as the consequence of its message-inherent news factors. But while responding to previous messages appears as a necessary element of discussing issues, quality news discussions also require listening to other discussants, expanding meaning, and learning about alternative points of view (e.g., Freelon, 2010). Therefore, investigating whether news discussions fulfill these criteria and if discussion factors support or limit these outcomes are important subjects for future research. Conclusion The current study extends the applicability of news value theory to media-stimulated interpersonal communication, shows that discussion factors help to explain why some user discussions on mass media websites are more interactive than others, and ultimately sheds light on their interpersonal nature. Despite the prominently displayed news items (as a specific manifestation of mass communication), some of the processes and contents of traditional interpersonal conversations about news and politics—such as asking questions, provoking others, providing new perspectives, or addressing specific discussants (e.g., Gamson, 1992)—can be effective in stimulating user-to-user interactions in online news discussions, too. Thus, the current study might have provided a rather general catalog of discussion factors that trigger user-to-user interactions and an important component for developing a theory that assesses the discussion value of online news items as a combination of the qualities of news items, published user comments, and discourse architectures. Acknowledgments The authors would like to thank the two anonymous reviewers for their thorough reading of an earlier version of this manuscript and for their highly constructive and helpful comments on how to improve the quality of this contribution. Parts of this research were supported by German Research Foundation Grant QU 215/3-1 to O.Q. Notes 1 " Such behavioral effects have been examined in other contexts, but mainly from the perspective of public opinion and not based on individual user comments (e.g., Yun & Park, 2011). 2 " In this study, the terms “response comments” and “feedback” are used interchangeably. 3 " Relevance is a relative and context-dependent construct (e.g., van Dijk, 1988). We understand relevance to be the result of a cognitive process by which media consumers assign some personal importance or meaningfulness to some aspects of a news item or a user comment (van Dijk, 1988). Put differently, “an input (…) is relevant to an individual when it connects with background information he has available to yield conclusions that matter to him” (Wilson & Sperber, 2004, p. 608). 4 " Users writing comments at least several times a week were counted as “regular” commenters, and users who wrote comments at least several times a month were counted as “occasional” commenters. 5 " Translations from the German transcripts were discussed with a bilingual native speaker (German and English) to best preserve the original meaning and tonality. 6 " As soon as interactions are introduced into the model, the main effects of the discussion factors, being part of the product terms of the interactions, cannot be interpreted as unconditioned effects anymore (Jaccard, 2001). Thus, the main effects of the discussion factors are reported in model I, and the interactions in model II. 7 " Another possible explanation can be derived from a “community perspective”: Although commenters can use pseudonyms on web platforms, there might be a community of people who regularly comment on a specific site. These users might know each other by means of their pseudonyms. As a result, they might also know that some users regularly write aggressive comments and that it is useless to interact with them. In contrast, there is a broad range of possible discussion spaces on Facebook within which users might interact only sporadically. Consequently, these users might to a lesser extent be able to anticipate whether it will make sense to discuss with a user who writes aggressive comments. References Alexa . ( 2013 ). Top sites in Germany . Retrieved from http://www.alexa.com Alonzo , M. , & Aiken , M. ( 2004 ). Flaming in electronic communication . Decision Support Systems , 36 , 205 – 213 . doi:10.1016/S0167-9236(02)00190-2. Google Scholar Crossref Search ADS WorldCat Anderson , A. A. , Brossard , D., Scheufele , D. A., Xenos , M. A., & Ladwig , P. ( 2013 ). The “nasty effect”: Online incivility and risk perceptions of emerging technologies . Journal of Computer-Mediated Communication Online first. doi:10.1111/jcc4.12009. OpenURL Placeholder Text WorldCat Boczkowski , P. J. , & Mitchelstein , E. ( 2012 ). How users take advantage of different forms of interactivity on online news sites: Clicking, e-mailing, and commenting . Human Communication Research , 38 , 1 – 22 . doi:10.1111/j.1468-2958.2011.01418.x. Google Scholar Crossref Search ADS WorldCat Bucy , E. P. ( 2004 ). Interactivity in society: Locating an elusive concept . The Information Society , 20 , 373 – 383 . doi:10.1080/01972240490508063. Google Scholar Crossref Search ADS WorldCat Dahlberg , L. ( 2001 ). Computer-mediated communication and the public sphere: A critical analysis . Journal of Computer-Mediated Communication , 7 . doi:10.1111/j.1083-6101.2001.tb00137.x. OpenURL Placeholder Text WorldCat Deutschmann , P. J. , & Danielson , W. A. ( 1960 ). Diffusion of knowledge of the major news story . Journalism Quarterly , 37 , 345 – 355 . Google Scholar Crossref Search ADS WorldCat Diakopoulos , N. A. , & Naaman , M. ( 2011 ). Towards quality discourse in online news comments. In CSCW '11 Proceedings of the ACM 2011 conference on Computer supported cooperative work (pp. 133–142). New York, NY: ACM. Retrieved from http://www.nickdiakopoulos.com/wp-content/uploads/2007/05/pr220-diakopoulos.pdf Eilders , C. ( 2006 ). News factors and news decisions: Theoretical and methodological advances in Germany . Communications , 31 , 5 – 24 . doi:10.1515/COMMUN.2006.002. Google Scholar Crossref Search ADS WorldCat Entman , R. M. ( 1993 ). Framing: Toward clarification of a fractured paradigm . Journal of Communication , 43 , 51 – 58 . doi:10.1111/j.1460-2466.1993.tb01304.x. Google Scholar Crossref Search ADS WorldCat Freelon , D. G. ( 2010 ). Analyzing online political discussion using three models of democratic communication . New Media & Society , 12 , 1172 – 1190 . doi:10.1177/1461444809357927. Google Scholar Crossref Search ADS WorldCat Galtung , J. , & Ruge , M. H. ( 1965 ). The structure of foreign news . Journal of Peace Research , 2 , 64 – 91 . doi:10.1177/002234336500200104. Google Scholar Crossref Search ADS WorldCat Gamson , W. A. ( 1992 ). Talking politics . Cambridge, MA : Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Herring , S. C. , Job-Sluder , K., Scheckler , R., & Barab , S. A. ( 2002 ). Searching for safety online: Managing “trolling” in a feminist forum . The Information Society , 18 , 371 – 383 . Google Scholar Crossref Search ADS WorldCat Jaccard , J. ( 2001 ). Interaction effects in logistic regression. Sage university papers: Series: Quantitative applications in the social sciences: Vol. 135 . Thousand Oaks, CA : Sage . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jones , Q. , & Rafaeli , S. ( 2000 ). Time to split, virtually: ‘Discourse architecture’ and ‘community building’ create vibrant virtual publics . Electronic Markets , 10 , 214 – 223 . Google Scholar Crossref Search ADS WorldCat Krippendorff , K. ( 2004 ). Reliability in content analysis . Human Communication Research , 30 , 411 – 433 . doi:10.1111/j.1468-2958.2004.tb00738.x. OpenURL Placeholder Text WorldCat Lee , E.-J. , & Jang , Y. J. ( 2010 ). What do others' reactions to news on internet portal sites tell us? Effects of presentation format and readers' need for cognition on reality perception . Communication Research , 37 , 825 – 846 . doi:10.1177/0093650210376189. Google Scholar Crossref Search ADS WorldCat Lindlof , T. R. , & Taylor , B. C. ( 2002 ). Qualitative communication research methods (2nd ed.). Thousand Oaks, CA : Sage . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lippmann , W. ( 1922 ). Public opinion . New York : Free Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lombard , M. , Snyder-Duch , J., & Bracken , C. C. ( 2002 ). Content analysis in mass communication: Assessment and reporting of intercoder reliability . Human Communication Research , 28 , 587 – 604 . doi:10.1111/j.1468-2958.2002.tb00826.x. Google Scholar Crossref Search ADS WorldCat Ng , E. W. J. , & Detenber , B. H. ( 2005 ). The impact of synchronicity and civility in online political discussions on perceptions and intentions to participate . Journal of Computer-Mediated Communication , 10 . doi:10.1111/j.1083-6101.2005.tb00252.x. OpenURL Placeholder Text WorldCat Östgaard , E. ( 1965 ). Factors influencing the flow of news . Journal of Peace Research , 2 , 39 – 63 . doi:10.1177/002234336500200103. Google Scholar Crossref Search ADS WorldCat Papacharissi , Z. ( 2004 ). Democracy online: Civility, politeness, and the democratic potential of online political discussion groups . New Media & Society , 6 , 259 – 284 . doi:10.1177/1461444804041444. Google Scholar Crossref Search ADS WorldCat Postmes , T. , Spears , R., & Lea , M. ( 1998 ). Breaching or building social boundaries?: SIDE-effects of computer-mediated communication . Communication Research , 25 , 689 – 715 . doi:10.1177/009365098025006006. Google Scholar Crossref Search ADS WorldCat Preece , J. ( 2001 ). Sociability and usability in online communities: Determining and measuring success . Behavior and Information Technology Journal , 20 , 1 – 15 . doi:10.1080/01449290110084683. OpenURL Placeholder Text WorldCat Price , V. , Nir , L., & Cappella , J. N. ( 2006 ). Normative and informational influences in online political discussions . Communication Theory , 16 , 47 – 74 . doi:10.1111/j.1468-2885.2006.00005.x. Google Scholar Crossref Search ADS WorldCat Purcell , K. , Rainie , L., Mitchell , A., Rosenstiel , T., & Olmstead , K. ( 2010 ). Understanding the participatory news consumer . Retrieved from http://www.pewinternet.org/˜/media//Files/Reports/2010/PIP_Understanding_the_Participatory_News_Consumer.pdf Quiring , O. ( 2009 ). What do users associate with ‘interactivity’? A qualitative study on user schemata . New Media & Society , 11 , 899 – 920 . doi:10.1177/1461444809336511. Google Scholar Crossref Search ADS WorldCat Rafaeli , S. , & Ariel , Y. ( 2007 ). Assessing interactivity in computer-mediated research. In A. N. Joinson, K. Y. A. McKenna, T. Postmes, & U.-D. Reips (Eds.), The Oxford handbook of Internet psychology (pp. 71 – 88 ). Oxford, England : Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rafaeli , S. , & Sudweeks , F. ( 1997 ). Networked interactivity . Journal of Computer-Mediated Communication , 2 . doi:10.1111/j.1083-6101.1997.tb00201.x. OpenURL Placeholder Text WorldCat Reich , Z. ( 2011 ). User comments: The transformation of participatory space. In J. B. Singer, A. Hermida, D. Domingo, A. Heinonen, S. Paulussen, T. Quandt, … M. Vujnovic (Eds.), Participatory journalism: Guarding open gates at online newspapers (pp. 96 – 117 ). Malden, MA : Wiley-Blackwell . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Ruiz , C. , Domingo , D., Micó , J. L., Díaz-Noci , J., Meso , K., & Masip , P. ( 2011 ). Public sphere 2.0? The democratic qualities of citizen debates in online newspapers . The International Journal of Press/Politics , 22 , 463 – 487 . doi:10.1177/1940161211415849. Google Scholar Crossref Search ADS WorldCat Schank , R. C. ( 1977 ). Rules and topics in conversation . Cognitive Science , 1 , 421 – 441 . doi:10.1207/s15516709cog0104_3. Google Scholar Crossref Search ADS WorldCat Schulz , W. F. ( 1982 ). News structure and people's awareness of political events . International Communication Gazette , 30 , 139 – 153 . doi:10.1177/001654928203000301. Google Scholar Crossref Search ADS WorldCat Shoemaker , P. J. ( 1996 ). Hardwired for news: Using biological and cultural evolution to explain the surveillance function . Journal of Communication , 46 ( 3 ), 32 – 47 . Google Scholar Crossref Search ADS WorldCat Shoemaker , P. J. , & Cohen , A. ( 2006 ). News around the world: Content, practitioners, and the public . New York, NY : Routledge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Singer , J. B. ( 2009 ). Separate spaces: Discourse about the 2007 Scottish elections on a national newspaper web site . The International Journal of Press/Politics , 14 , 477 – 496 . doi:10.1177/1940161209336659. Google Scholar Crossref Search ADS WorldCat Springer , N. , & Pfaffinger , C. ( 2012 , May). Why users comment online news and why they don't. Paper presented at the 62nd Annual Conference of the International Communication Association, Phoenix, AZ. Staab , J. F. ( 1990 ). The role of news factors in news selection: A theoretical reconsideration . European Journal of Communication , 5 , 423 – 443 . Google Scholar Crossref Search ADS WorldCat Strauss , A. L. , & Corbin , J. M. ( 1990 ). Basics of qualitative research: Grounded theory procedures and techniques (17th ed.). Newbury Park, CA : Sage . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Stromer-Galley , J. ( 2003 ). Diversity and political conversations on the internet: Users' perspectives . Journal of Computer-Mediated Communication , 8 . doi:10.1111/j.1083-6101.2003.tb00215.x. OpenURL Placeholder Text WorldCat Stromer-Galley , J. ( 2004 ). Interactivity-as-product and interactivity-as-process . The Information Society , 20 , 391 – 394 . doi:10.1080/01972240490508081. Google Scholar Crossref Search ADS WorldCat Stromer-Galley , J. ( 2007 ). Measuring deliberation's content. A coding scheme . Journal of Public Deliberation , 3 , 1 – 35 . OpenURL Placeholder Text WorldCat Stromer-Galley , J. , & Martinson , A. M. ( 2009 ). Coherence in political computer-mediated communication: Analyzing topic relevance and drift in chat . Discourse & Communication , 3 , 195 – 216 . doi:10.1177/1750481309102452. Google Scholar Crossref Search ADS WorldCat Suler , J. ( 2004 ). The online disinhibition effect . Cyberpsychology & Behavior , 7 , 321 – 326 . doi:10.1089/1094931041291295. Google Scholar Crossref Search ADS WorldCat Sundar , S. S. ( 2004 ). Theorizing interactivity's effects . The Information Society , 20 , 385 – 389 . doi:10.1080/01972240490508072. Google Scholar Crossref Search ADS WorldCat Tong , A. , Sainsbury , P., & Craig , J. ( 2007 ). Consolidated criteria for reporting qualitative research (COREQ): A 32-item checklist for interviews and focus groups . International Journal for Quality in Health Care , 19 , 349 – 357 . doi:10.1093/intqhc/mzm042. Google Scholar Crossref Search ADS PubMed WorldCat Trepte , S. , & Reinecke , L. ( 2011 ). The social web as a shelter for privacy and authentic living. In S. Trepte & L. Reinecke (Eds.), Privacy online. Perspectives on privacy and self-disclosure in the social web (pp. 61 – 73 ). Berlin : Springer . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Turner , T. C. , Smith , M. A., Fisher , D., & Welser , H. T. ( 2005 ). Picturing usenet: Mapping computer-mediated collective action . Journal of Computer-Mediated Communication , 10 . doi:10.1111/j.1083-6101.2005.tb00270.x/full. OpenURL Placeholder Text WorldCat van Dijk , T. A. ( 1988 ). News as discourse. In Communication (2nd ed.). Hillsdale, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Walther , J. B. , & Jang , J.-w. ( 2012 ). Communication processes in participatory websites . Journal of Computer-Mediated Communication , 18 , 2 – 15 . doi:10.1111/j.1083-6101.2012.01592.x. Google Scholar Crossref Search ADS WorldCat Warren , C. N. ( 1934 ). Modern news reporting . Madison, Wis .: Harper & Brothers . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Weber , P. ( 2013 ). Discussions in the comments section: Factors influencing participation and interactivity in online newspapers' reader comments . New Media & Society Online first. OpenURL Placeholder Text WorldCat Willemsen , L. M. , Neijens , P. C., Bronner , F., & de Ridder , J. A. ( 2011 ). “Highly recommended!” The content characteristics and perceived usefulness of online consumer reviews . Journal of Computer-Mediated Communication , 17 , 19 – 38 . doi:10.1111/j.1083-6101.2011.01551.x. Google Scholar Crossref Search ADS WorldCat Wilson , D. , & Sperber , D. ( 2004 ). Relevance theory. In L. R. Horn & G. L. Ward (Eds.), Blackwell handbooks in linguistics: Vol. 16. The handbook of pragmatics (pp. 607 – 632 ). Malden, MA : Blackwell . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wright , S. , & Street , J. ( 2007 ). Democracy, deliberation and design: The case of online discussion forums . New Media & Society , 9 , 849 – 869 . doi:10.1177/1461444807081230. Google Scholar Crossref Search ADS WorldCat Yun , G. W. , & Park , S.-Y. ( 2011 ). Selective posting: Willingness to post a message online . Journal of Computer-Mediated Communication , 16 , 201 – 227 . doi:10.1111/j.1083-6101.2010.01533.x. Google Scholar Crossref Search ADS WorldCat Ziegele , M. , & Quiring , O. ( 2013 ). Conceptualizing online discussion value: A multidimensional framework for analyzing user comments on mass-media websites . Communication Yearbook , 37 , 125 – 154 . OpenURL Placeholder Text WorldCat © 2014 International Communication Association
TI - What Creates Interactivity in Online News Discussions? An Exploratory Analysis of Discussion Factors in User Comments on News Items
JO - Journal of Communication
DO - 10.1111/jcom.12123
DA - 2014-12-01
UR - https://www.deepdyve.com/lp/oxford-university-press/what-creates-interactivity-in-online-news-discussions-an-exploratory-dTKVOBnXi3
SP - 1111
VL - 64
IS - 6
DP - DeepDyve
ER -