Online Newspapers’ Readers’ Comments - Democratic Conversation Platforms or Virtual Soapboxes?

Online discussions between citizens have been subject to scholarly debate and several empirical studies for a substantial period of time. Scholars have envisioned these as potential venues for citizens’ deliberation. Often, however, empirical findings do not support these notions. The quality of discussion is often not up to the standards of truly deliberative discussions. Nonetheless, there are strains of evidence suggesting that online discussion venues applying facilitators, fixed topics and actively enforcing discussion rules, may avoid many of the problems with discussion quality. This article therefore examines whether readers’ comments on a Finnish newspaper’s website resemble democratic conversations. During the course of three weeks in spring 2010, we content analyze 300 reader comments. Our findings indicate that although there are strains of democratic conversations to be found in online reader comments, there are also many instances in which the opposite holds true.


Introduction
The democratic value of public discussions between citizens is a topic which has recently experienced a surge within democratic theory (Chambers, 2003).Some scholars (e.g.Dryzek, 2000: 1) even claim that "the essence of democracy itself is now widely taken to be deliberation, as opposed to voting and interest aggregation".This is, of course, an idea which is far from novel; the thoughts of theorists such as Habermas (1989), Dahl (1989) and Barber (1984) serve as inspiration for the contemporary focus on deliberation.That which distinguishes the current focus on citizen deliberation from earlier "turns", however, are the efforts made to convert theory into practice through various forms of mini-publics (Dryzek & Goodin, 2006).Albeit this article does not focus on deliberation nor mini-publics per se, it nonetheless draws on deliberative theory and focuses on conversations between citizens on the internet.Bearing in mind that democratic conversations require something of a face-to-face society, i.e. groups of a manageable size allowing for talking and reasoning (Fishkin & Laslet,t 2003), the internet appears to have many favorable conditions for democratic conversations.On this note, Wright and Street (2007: 72) mention that "it is suggested that the technical characteristics of the internet […] have led to the creation of a 'virtual world' that establishes the conditions for deliberative democracy".
However, conditions do not always entail that promises are fulfilled.Deliberative theory has several conditions that citizen conversations need to satisfy in order to be considered democratic conversations (see Barber, 1984;Dahlberg, 2001;2004;Stromer-Galley, 2007).In this light, many studies of online discussions between citizens have not readily found support (e.g.Ferber et al., 2006;Jankowski & Van Os, 2004;Papacharissi, 2002;Strandberg, 2005;2008;Witschge, 2004).Nonetheless, the empirical data gathered from the efforts made concerning off-line deliberation events (e.g.Fishkin et al., 2002) and, to some extent, regarding online equivalents (e.g.Grönlund et al., 2009;Strandberg & Grönlund, 2012) shed light on an interesting observation; citizen discussion venues designed to be democratic have a better chance of achieving conversations of the type perceived beneficial for democracy (Coleman & Moss, 2012: 6;Wright & Street, 2007: 850).
Drawing on that observation, this article focuses on a venue for citizen discussions which, with some exceptions (e.g.Friedman, 2011;Kohn & Neiger, 2007;Diaz Noci et al., 2010;Ruiz et al., 2011), has not received much scholarly attention in the context of citizen deliberations: reader comments on online newspaper stories.Concerning this Ruiz and colleagues (Ruiz et al., 2011: 464) stress that: " […] it is time to assess whether online newspapers provide a new incarnation of the public sphere, becoming the digital cafés of a Public Sphere 2.0".
In this article, we ask to what extent citizen discussions in reader comments constitute democratic conversations.The article thus takes an explorative and descriptive approach, and it should be stressed that we do not have ex ante facto normative expectations that online reader comments should fulfill all the criteria of democratic conversations.The empirical analysis employs content analysis and focuses on comments to a Finnish newspaper's online edition in spring 2010 (further description in data and methods section).

Democratic Discussions
Which are the criteria of a democratically beneficial communicative process?While there are many variations in terms, many core elements similar to each other can be found in the reasoning of most scholars regarding the indicators of democratic conversations (e.g.Barber, 1984;Dahlberg, 2001;Jankowski & Van Selm, 2000;Stromer-Galley, 2002).These should ideally have equality among discussants (inclusion), contain rationality, relevance to the topic, sincerity, reciprocity, and interaction.Moreover, the discussants must show each other mutual respect.
The notion of the public sphere entails that citizen gather to discuss the issues of the day (cf.Habermas, 1989), and some have argued that modern mass media can broaden this sphere since there are always practical difficulties associated with enabling large numbers of citizens to be physically co-present.This view has received a significant surge in scholarly attention with the advent of the internet, which by many is regarded as a potential virtual public sphere even superior to traditional mass media (see for example Papacharissi 2002;Ruiz et al. 2011;Singer, 2009).

ICTs and Democratic Discussions
The democratic potential of the internet lies in its diversity as a medium and also in its interactivity.Compared to traditional media, the internet has several benefits including speed, low costs of usage, the possibility of easily communicating multimedia and allowing for publication without interference from editors (Strandberg, 2005).The internet thus has the ability to allow citizens to become more creatively involved in political processes by offering them accessible communication tools (e.g.Barber, 1984;Dahl, 1989).These tools nowadays include writing postings on an online discussion forum, blogging, commenting on news stories, sharing content, tweeting, joining or starting a Facebook-group.
Online citizens' discussions can, and have, been viewed as electronic versions of public spheres (Ruiz et al., 2011: 464).Moreover, they can be held easily regardless of physical location (Papacharissi 2002), they encourage rational argumentation (Papacharissi, 2004), they promote political discussion between citizens (Nisbet & Scheufele, 2004) and are, at best, characterized by a will to come up with the best argument as well as finding shortcomings in other debaters arguments (Graham, 2009).The empirical results concerning the quality of online discussion arguably are, however, inconclusive.Even though the discussion fulfills the criteria of democratic conversations in some cases (see Stromer-Galley, 2002;Jensen, 2003;Graham, 2009), they are often found to be far from fulfilling the criteria of democratic conversations in other cases (Wilhelm, 2000;Strandberg, 2005;2008;Ferber et al., 2006).

Reader comments on online newspaper stories
Newspapers, and especially their public sections and letters to the editor, have traditionally been a very visible forum for debate and influencing public opinion.According to Ruiz et al. (2011: 464), they can even be considered the most popular form of audience participation.Readers' comments on online newspaper articles represent a relatively new forum for discussion between citizens and possess -at least in theory -the necessary characteristics to function as venues for democratic conversations.Readers' online commenting is an interactive format allowing readers of an online newspaper to express their views and discuss published articles.Usually readers are able to post their comments directly in attachment to the article, often in a comment-field beneath the text.The characteristics of such reader comments -immediate publishing, large amounts of space and minimal censorship -open up possibilities for readers to participate in a way letters to the editor or the public section never can due to limited space in print newspapers.Reader comments enable citizens to discuss topics in context since the article they are connected to marks the subject of discussion (Manosevitch & Walker, 2009).Domingo (2008: 694), similarly, describes reader comments as "micro forums attached to news".This means that reader comments might be more visible than other independent online forums and at the same time facilitate an open and public discussion.
Correspondingly to the scholarly discussion concerning online forums in general, the view on whether the democratic potential of citizen discussion could be fulfilled in the form of newspapers' readers' comments is rather divided.Some are adamant that readers' comments will change political communication in a positive manner (see for example Hecht, 2003;Manosevitch & Walker, 2009;Schuth et al., 2007), while other scholars express concerns about the low democratic quality and lack of deliberation in readers' comments (Kohn & Nieger, 2007;Skjerdahl, 2008;Hedman, 2009) as well as the little interest shown by the public for submitting comments (Bergström, 2008).
Regarding empirical research findings, Nagar (2009) and Hedman (2009), for instance, have found that many newspaper editors are clearly concerned about the content and quality of readers' comments.In a rhetorical analysis of Israeli readers' comments, Kohn and Neiger (2007) noticed that the comments reflected a violent and aggressive discussion, while simultaneously containing verbal finesse.Hecht's conclusion (quoted in Nagar, 2009) was that readers' comments constitute a spontaneous discussion on current affairs and differ from blog comments by gaining more authority in the eyes of the citizens.Manosevitch & Walker (2009) state that readers' comments on editorials pay close attention to the editorial itself and to other comments and may become a democratic tool for citizens.Still, opposite findings are presented by Diaz Noci et al. ( 2010) who noticed that there is a lack of respect, diverse perspectives and mature arguments.Hence they propose that "comments in news are not fostering a democratic dialogue".Diaz Noci et al. also imply that human moderation before the publication of comments is the price to pay if the goal is to guarantee a democratic debate.In a cross-national study of online comments to stories from various media systems in the leading dailies in the US, UK, France, Spain and Italy, Ruiz et al. (2011) assess the democratic quality of the readers' comments.These scholars found that commenting to online newspapers can be divided into two types: i) communities of debates and ii) homogeneous communities.In the former, they remark that opposite views to those of the majority are welcomed and foster debates whereas, in the latter type, readers' comments tend to be a "collective reproduction of the same positions".Friedman (2011: 13), focusing on readers' comments on the Israeli-Palestinian conflict in two online newspapers, arrives at rather similar conclusions in as much as he argues that online reader commenting is best described as a mix of "critical-rational debate, extremist demagoguery, friendly teasing, exhibitionism, and emotional pleas".
To summarize this section, we draw attention to the fact that there is still much theoretical uncertainty and empirical inconclusiveness regarding the democratic potential of online readers' comments.As Ruiz et al. (2011: 464) remark "there has been little scholarly attention toward the nature of these conversations [readers' comments, authors' note] and their implications of democracy".Thus, we are interested in further exploring the question of to what degree and how these online "micro forums attached to news" (Domingo, 2008: 694) fulfill the ideals of deliberative democratic conversations.Through analyzing a newspaper in Finland we also hope to provide evidence from a communicative context -the democratic-corporatist media systems (Hallin & Mancini, 2004) -which has been highlighted as a context in need of more study (e.g.Ruiz et al., 2011).

Approach, data and methods
In exploring to what degree and how online readers' comments fulfill the ideals of deliberative democratic conversations we will conduct an ideal-type analysis of sort (see Esaiasson et al., 2003: 155) through examining the extent to which debates in newspapers' readers' comments fulfill the normative standards of democratic conversations -i.e. the ideal-type of such conversations.Constructing such an ideal-type is not necessarily an easy task as it requires 'translating' normative assessments into measurable indicators (see Downey et al., 2012).Thankfully, though, a lot of work has already been done regarding this 'translation' both by scholars applying deliberative principles in practice through mini-publics (e.g Dryzek & Goodin, 2006) and by those who have studied the democratic quality of online discussions (e.g.Dahlberg, 2001;Jankowski & Van Selm, 2000).Although we are aware that additional conditions for deliberation have been used in the literature (e.g.Trénel, 2004;Graham, 2009;Stromer-Galley, 2007), we focus on four areas of interest in an attempt to concentrate on the most crucial conditions for deliberation: (i) the rationality of the debates (ii) the relevance to the topic of the debates (iii) the reciprocity of the debates (iv) the degree of politeness and respect in the debates These areas can, however, be measured and operationalized in various manners (e.g.Graham, 2009;Jensen, 2003;Manosevitch & Walker, 2009;Steenberger et al., 2003).As a common denominator, though, quantitative content analysis has been applied.In finding an appropriate coding scheme for our study, we conducted a pilot test in which existing code schemes were tested on a small portion of our data.Based on the pilot test, it became evident that an adapted scheme using elements from several schemes needed to be constructed 1 .The reliability of our scheme was tested using the formula suggested by Holsti (1969) and showed a highly satisfying .93degree of intra-reliability.Even though only one coder was used, we checked the scheme's inter-coder reliability which amounted to .83 using the same formula.This adapted scheme measures the four areas of interest using several different indicators adapted from the literature (Table 1): Peering into the rationality of the debates we look at several alternative indicators: justifications of claims (Steenberger et al., 2003), reasoning of claims according to argument type (Graham, 2009), validations of claims, as well as the type of evidence used to support claims (Graham, 2009;Jensen, 2003).These are means of assessing to what extent debaters do not merely state personal opinions and instead make efforts to logically and, by argumentation or use of references, rationally defend and elaborate their claims.Relevance, i.e. staying on topic, is probably one of the most clear-cut aspects of discursive quality in any form of debate (Schneider, 1997;Jensen, 2003;Stromer-Galley, 2007;Graham, 2009).Concerning comments to news articles, we define the topic as the subject of the news story.Reciprocity is very close to the essence of democratic conversations: "Strong democratic talk, then, always involves listening as well as speaking, feeling as well as thinking, and acting as well as reflecting" (Barber, 1984: 178).In order to gauge this aspect, we analyze three indicators in the comments (cf.Ruiz et al., 2011); the direction of interaction; the occurrence of disagreement/agreement; and the type of reciprocity.The final condition, politeness and respect, is both a precondition for the democratic quality of conversation and at the same time signs of quality in such conversations (Papacharissi, 2004; see also e.g.Graham, 2009;Jensen, 2003).We measure this condition with three indicators in comments; tone, degree of incivility, and the degree of impoliteness.
We do acknowledge, however, that these conditions and especially some of their indicators are somewhat interrelated and overlapping in some cases.Still, for the purpose of this article, in which we want to assess the democratic quality of online readers' comments (a rather unexplored area), we want to keep an open mind and test several ways of measuring aspects of democratic quality.In carrying out this task, a significant concern is how one can determine whether the discussions held in the article comments satisfy the normative conditions for deliberation.Graham (2009: 53-54) also acknowledges that researchers never explicitly define when the normative conditions are fulfilled.What are, for instance, the cut-off points between a low-quality and a high-quality discussion?Wright (2011: 12) likewise acknowledges this problem and advises researchers to adopt a more flexible approach towards online discussions.According to Wright (ibid: 6-7), researchers simply often create undue expectations before evaluating their empirical findings, which can lead them to conclusions that are too pessimistic.In conducting our ideal-type analysis, thus, we also briefly provide some comparisons to findings from corresponding studies in order to balance our assessments on the normative conditions.Our data derives from readers' comments to news stories in the online version of a regional Finnish newspaper, Vasabladet, in the city of Vaasa.Naturally, focusing on a single case has its drawbacks and limitations (Peters, 1998: 5-6).The Vasabladet newspaper is not, for example, a leading Finnish newspaper.Ruiz et al. (2011: 483) do, however, stress the value of looking beyond leading national newspapers in analyses of online comments in news (cf.Downey et al., 2012: 340).Moreover, the Vasabladet newspaper is not without a large online readership.The average number of unique visitors per week is 38,031 (Vasabladet, 2010).The average age of the online readers is 42 years and almost Online Newspapers' Readers' Comments -Democratic Conversation Platforms or Virtual Soapboxes? .Kim Strandberg, Janne Berg two thirds of the online readers are male2 .The site requires registration -albeit it is possible to have a nickname nonetheless -before readers are allowed to post comments.The website uses outsourced post-moderation of the readers' comments (ibid.)Furthermore, there are other circumstances which make a Finnish online newspaper an interesting case.Firstly, Finland has a very long tradition of having a high internet penetration rate; online news is now among the media channels most frequently used by Finnish citizens generally, including when reading about political matters (Strandberg, 2012;Statistics Finland, 2010).Finland also has the third largest newspaper readership in the world 3 .
The data for our study was collected between March 29th and April 18th 2010.During this period a total of 305 news stories were posted online on the Vasabladet website.The news stories whose discussions we analyzed concerned various topics -albeit mostly societal-or political matters (about 73% of all stories).Our motivation for not limiting our material to specifically political news topics is the research of, for instance, Coleman and Blumler (2009) and Graham (2009), who all have argued and shown that the democratic potential of forums with non-political purposes should not be neglected (cf.Wright, 2011).Of the 305 news stories covered in this research, 2,313 had received comments in total, yielding an average of 7.58 comments per story.The comments analyzed were randomly selected among comments to news stories which had received at least 10 comments.This threshold was needed in order to assure that the comments analyzed were part of lengthier discussions in which actual debate could have taken place.The total number of such articles with at least 10 comments amounted to 91 (30% of the total 305 articles).In order to keep the number of comments manageable given the in-depth coding scheme, we then randomly selected 300 comments for analysis from the 91 news stories with at least 10 comments attached to them.The average length of these 300 comments was 232 characters (including spaces) with a very high standard deviation of 206 (comments' length ranging from 3 characters to 1348 characters).The number of unique pseudonyms posting these comments was 212, making the average of comments per pseudonym 1.4, with a standard deviation of 2.6.The maximum amount of comments posted by a pseudonym was nine, while 37 comments had a blank pseudonym.Concerning this, however, it should be noted that it was entirely possible for a citizen to change nicknames every time he or she posted a comment, so the average of comments per pseudonym is only to be regarded as a rough estimate.It is also noteworthy that only three nicknames posted more than four comments each (cf.Ruiz et al., 2011: 475-476).Almost all, i.e. 96% of the citizens posting comments did so using nicknames which were entirely anonymous.

Findings
We present our findings according to our four areas of focus and their corresponding indicators.Throughout this presentation we also assess our findings in light of the deliberative ideals from which they have been derived.As stated earlier, when applicable, we also briefly compare some of our findings to those of other studies.

Rationality of discussions
In our census of 300 comments, 83% (250 comments) contained some form of claim on the part of the citizen posting the comment.In analyzing the rationality of these claims, then, we start by looking at how claims were justified (Table 2).Table 2 shows that a majority of claims (57.2 %) either lacked justification or had a poor justification.Nonetheless, 43% of the claims were substantiated in a satisfying manner (a qualified or sophisticated justification), entailing that the comment in question contained a logically sound conclusion.Judging this finding in normative terms, we would assess that the justification of comments was satisfactory but not ideal, in as much as a large share of comments lacked sufficient justification and only a few comments had sophisticated justification.These findings are, however, quite similar to findings regarding the level of justification in other contexts (Himmelroos, 2010;Lord & Tamvaki, 2011;Spröndli, 2003).Only studies of parliamentary debates in Switzerland and the UK have demonstrated a significantly higher level of justification (Steiner et al., 2004).
Looking at how claims are reasoned by studying the argument type supporting the claim made, we note that using this definition 70% (175 comments out of 250) of the claims made in comments were reasoned, amounting to 58% of all comments (175 of 300).Again, this indicator of rationality is met to a satisfactory degree but still has plenty of room for improvement.Our finding is also slightly lower than in some other studies (e.g.Graham, 2009;Zhou et al., 2008) but in line with Friedman's study (2011).The distributions of reasoning of claims according to type are presented in Table 3.As seen in Table 3, most of the claims analyzed were reasoned through alternative arguments, i.e. claims made without trying to either contradict or challenge opposing claims.Only 20% of the claims provided counterevidence in order to contradict or challenge opposing claims.From a normative perspective one could argue that providing such evidence would be the most rational type of motivation.Consequently, the reasoning of claims in our study is rather far from the deliberative ideal of rational discourse.

n %
In Table 4, we assess the next indicator of rationality; the use of internal and external validations of claims.In Table 4, we see that the majority of the validations were internal validations.This means that the citizen validated a claim using only his/her own views and values as a basis.External validation, i.e. when the citizen uses information from external sources (facts, figures for example), was less common and used only in 12.7% of the comments.Comments containing both internal and external validations were rare; these amounted to only 7.7% of all comments.Clearly, thus, the use of validations in the comments was far from the ideals of truly democratic conversations.If such discussion ought to be a "critical and rational discourse" (Graham, 2012: 33), a majority of validations should not be based on peoples' own values and beliefs.

No
Scrutinizing one final indicator of rationality, the supporting of claims using references or evidence, an initial overview (not reported in a table) shows that some sort of reference to support a claim or opinion was used in 52.3% (157) of the comments.In total, different levels or types of evidence were used 189 times (a comment could contain more than one level of evidence).However, as evident in Table 5, facts or sources were only referenced in substantiating 19.6% of the comments (37 out of 300).Personal experience was referenced equally as much and drawing parallels or stating some form of example was used most frequently.The use of parallels or comparisons was the most popular level of evidence and was applied in 34.9% of the comments.In a sense, thus, the supporting of claims with evidence was satisfactory when considered from the idealtype perspective.The 'best' finding here would have been that most claims had been supported with facts or sources, which evidently was not the case.However, the 'worst possible' result of all claims being supported with merely personal experience as the evidence type was equally as far from the actual case.Note: A comment may have several levels of evidence.130 comments had only one level of evidence, 23 comments had two levels, three comments had three levels and one comment contained all four levels of evidence.
Wrapping up the section on rationality, then, it appears that the comments on the Vasabladet website's news stories were not readily rational; the most rational categories in our variables were infrequently found for all but the justification indicator.The brief comparisons to other studies, however, would suggest that this normatively discouraging finding is not the reason for being overly pessimistic about our Finnish case.

Relevance of discussions
Article comments are often accused of being off-topic and irrelevant (e.g.Rieder, 2010).However, in the 300 comments we analyzed we found that most comments were either directly referenced to the topic (47.3%) or were statements or clarifications, which related to the topic of the article indirectly (28.7%).Meta-comments (comments mainly reflecting on the debate) were rare (10.3%).Irrelevant comments were also uncommon (13.7%).Overall, thus, the normative criterion of relevance is mostly fulfilled since three out of four comments related to the topic.This finding also echoes those found in most other studies of online readers' comments (e.g.Graham, 2009;Jensen, 2003;Diaz Noci et al., 2010;Ruiz et al., 2010) and was clearly more positive than the findings in both Schneider's (1997) and Stromer-Galley's (2007) studies.

The reciprocity of the debates
Continuing with the analysis of the reciprocity of the debates, we firstly take a glance at the direction of the comments made -i.e. at whom or what the comment is aimed.Our findings concerning this indicator show that the majority of the direction of statements in comments (N=388 since a comment can be aimed at several targets) were aimed at the news story (60.3%), and almost all of the remaining instances of directed statements (37.6%) were aimed at other comments.Hence, in light of this indicator of reciprocity, the finding is rather satisfactory from a normative point of view, although one would ideally have expected an even higher share of directed claims aimed at other comments.Peering further into the mechanisms of this reciprocity, we look at the degree of disagreement and agreement the debates (Table 6).Of all 300 comments, 59.3% (178 comments) did not contain any signs of agreement nor disagreement with other citizens' statements, meaning that the remaining 40.7% (122 out of 300) of the comments showed signs of interaction with other comments.The interaction with other comments was hence slightly higher using this indicator than the measure of directions of claims applied earlier.Nonetheless, among the comments which had interaction in some form, disagreement with a claim made by another commenter was the predominant type (35% of all comments) rather than agreement (9%).Whether agreement is closer to the ideal of democratic conversations than disagreement is hard to judge.Although it is arguably more unpleasant than agreement, disagreement can be seen as a sign of a heterogeneous discussion climate where a broad range of perspectives are considered (Stromer-Galley & Muhlberger, 2009: 175).Nonetheless, it is noteworthy that our findings stand in contrast to that of a previous deliberation experiment where agreement was more common than disagreement (Stromer-Galley & Muhlberger, 2009).
Turning to our final indicator of reciprocity, then, we shed further light on the type of reciprocity.Concerning this measure we find, rather remarkably, that a debater was convinced by the arguments made in a previous comment in only one comment among all the comments showing reciprocity indicators (N=122).In three quarters (73.8%) of the comments containing interaction, comments made by other citizens are merely noted without further acknowledgement of the merits of their claims.Additionally, in one quarter (25.4%) of the reciprocal comments, debaters make negative remarks on the comments of other debaters.In light of this indicator of reciprocity, thus, our findings are very far from the ideal of truly democratic conversations, since it appears that the interaction with other debaters' comments is quite shallow, to say the least.Judging from Jensen's findings (2003), where both lower amounts of negative reactions and a higher frequency of people being persuaded by others was found, our result is also put in an unfavorable light.
Considering the condition of reciprocity as a whole, then, we would tentatively state that these online debates were quite far from the normative ideals derived from deliberative theory.None of the three indicators used here revealed readily encouraging findings, and the findings of the analysis of the type of reciprocity were rather abysmal when judged against the ideals.

The degree of politeness and respect in the debates
In this final section of our analysis, we peer into this last of the four focus areasthe degree of politeness and respect -by assessing three indicators; the extent of impoliteness, the degree of incivility and the tone of the comments.Regarding the two former of these indicators, our findings are encouraging from the normative standpoint; almost all comments were civil (98.7%) and polite (85.3%).This finding could of course be inflated due to the post-moderation used by the Vasabladet-newspaper which might be disproportionally reducing impolite and rude comments (cf.Ruiz et al. 2011: 482-483).Nonetheless, our findings appear similar to those of other studies (c.f.Papacharissi, 2004;Ruiz et al., 2011;Diaz Noci et al., 2010).
Looking at the final indicator in Table 7, the tone of the comments, we find that there is a rather even split between comments being negative (42.3 %) and being neutral (41.7 %) or factual (14.0%).Hence, the tone of the comments on the Vasabladet website was quite satisfying from a normative standpoint, although the significant share of negative comments is noteworthy.In comparison to Jensen's (2003) study, the readers' comments analyzed here are less hateful, albeit somewhat more negative in tone and less respectful.Altogether, the normative condition of politeness and respect is arguably rather close to the ideals of truly democratic conversations.It is mainly the high degree of negativity found upon our analysis of the tone of the comments which detracts from the normative ideal.

Conclusions
As we stated in the outset of this article, we did not hold a priori expectations that normative conditions for democratic speech should be entirely fulfilled in article comments on newspaper websites.Rather, we used these normative conditions in our analysis as an ideal type -a yardstick of sort -towards which we could evaluate the extent of the democratic quality of the discussions held in the online articles' comments.Judging from our findings, then, how did the online comments on the Vasabladet website generally measure on this "normative yardstick"?
Generally, the answer provided by our study to the above question is both positive and negative.Positive signs can be seen in our findings concerning one aspect of rationality (justifications), and also relating to relevance of comments and politeness and respect.Negative, and less-than-ideal, findings are evident in light of most aspects of rationality and also regarding reciprocity.Evidently, thus, judged in light of normative ideals, our empirical observations are relatively far off the mark.However, in light of the brief posthoc comparison of our results to other studies' findings, it is fair to say that the reader comments analyzed here did not generally deviate more from ideal-speech than findings in other communicative contexts.A rather accurate assessment of discussions in online reader comments is that they sometimes function as the "digital cafés of a public sphere 2.0" (Ruiz et al. 2011, 464); but just as often, if not even more frequently, they do not.
Looking beyond empirical findings, there are also some methodological observations brought about during the process of carrying out our analysis which merit some attention.One such observation pertains to the suitability of using criteria derived from a normative theory in the specific context and type of data analyzed here.Even though we adapted a scheme, drawing on the operationalizations of deliberative criteria used by several different scholars in various contexts, we still found it rather troublesome to apply it in the actual analysis.Perhaps the criteria for democratic conversations are best suited in a context more readily designed for deliberation, where it is more reasonable to expect that democratic conversations should take place such as in deliberative minipublics (Dryzek & Goodin, 2006) or parliamentary debates (Steiner et al., 2004).After all, the newspapers providing the opportunity to comment on their online stories often do so with business-driven motives (see Ruiz et al., 2011: 464) and not with the Habermasian ideal speech situation in mind.Naturally, though, drawing on the criteria of the ideal speech situation in analyses of online speech in any context is a good starting point, although we ex ante facto agree with Wright's assessment (2011: 6-7) that this starting point might create undue expectations a priori, which can lead to conclusions that are too pessimistic.We would, nonetheless, hope that future studies seek to adapt the coding schemes even further than we have done here, so that the one-size-fits-all approach to analyzing democratic conversations can be further refined (cf.Downey et al., 2012).

Table 1 .
Areas of focus and their indicators.Note: See appendix for operationalizations Online Newspapers' Readers' Comments -Democratic Conversation Platforms or Virtual Soapboxes? .Kim Strandberg, Janne Berg

Table 2 .
Type of justification in comments containing at least one claim

Table 3 .
The reasoning of claims by argument type

Table 5 .
Level of evidence

Table 6 .
Agreement and disagreementNote: one comment can have instances of both disagreement and agreement with other people's comments.Hence, the categories are not regarded as mutually exclusive in one comment.