NEW VOICES IN THE PUBLIC SPHERE: A COMPARATIVE ANALYSIS OF INTERPERSONAL AND jennifer ONLINE POLITICAL TALK stromer-galley Abstract Political conversation for many people is a taboo activity, particularly with acquaintances or strangers. Online, there are a wealth of political conversation spaces, designed for acquaintances and strangers to interact. The question is are there people talking politics online who do not do so face-to-face. This essay presents findings on people's reported political conversation behaviour online and offline from secondary survey analysis of a research project studying the effects of political deliberation. The survey analysis suggests that there are people who talk politics online who do not do so in face-to-face situations, and they are categorically different than those who do so face-to-face. The Internet may provide a new context for political conversation for those who would not normally engage in face-to-face political conversations, thus bringing new voices into the public sphere. Jennifer Stromer-Galley is a Ph.D. candidate at Annenberg School for Communication, University of Pennsylvania, email: jstromer@asc.upenn.edu. Communication is the vehicle for politics. Who communicates, how they communicate, in what context, and to what effect are important considerations political scholars investigate. Many focus on elites who drive much of the political conversation, such as journalists, politicians, and political experts (Hart 1987; Jamieson 1988; Jamieson 1996; Katz and Lazersfeld 1960; Mutz 1998; Page 1996). More recently, effort has focused on citizens in deliberative forums, talking politics in structured ways, and with purposive outcomes (Fishkin 1995). These research efforts are important, but what is less studied are the organic, unstructured political conversations that occur spontaneously among non-elites. Huckfeldt and Sprague (1995) identified in the mid-90s that there is surprisingly little that we knew about "information received through social interaction, the characteristics of the relationships through which political information is conveyed, and the types of people who serve as political communicators" (p. 108); this observation is still relevant. Across the Internet, one can find spaces designed specifically for political conversation. In other spaces, designed for other kinds of conversations, political conversation naturally springs. These conversations - easily observable, readily join-able for anyone interested in doing so - offer researchers an intriguing opportunity to watch unstructured, political conversations to see who participates, what the content of the conversation consists of (Davis 1999; Davis and Owen 1998; Hill and Hughes 1998), and to what possible effects (Jankowski and Van Selm 2000; Streck 1998) . In the area of computer-mediated communication, there is a growing body of studies of the observed conversations online. Yet, there is much we do not know. One of the most basic questions that is not yet answered is: Why do people talk politics online? Face-to-face, interpersonal political conversation exists, and presumably people have opportunities to engage others offline about politics as they do online. So, why do they choose to do so online? Are the people who talk online the same people doing so offline? What motives are satisfied by such conversations? In this essay, I present findings that begin to address these questions. In the pages that follow, I map out literature on political conversation off and online. I then present findings that there are people who report online political talk who report no in-person political conversation partners. I also present findings from logistic regression models that suggest characteristic differences between those who talk with friends and family, with acquaintances, and online. I offer some initial observations into why they might be doing so. I then turn to public sphere theory and offer some ideas of how the Internet conversation context changes our notions of the traditional public sphere. Using Internet for Political Conversation Political conversation options on the Internet are numerous, and by political conversation I mean conversation on current events, social issues, public policy, political campaigns, and government. Many news and current events sites in the United States, such as CNN.com, the New York Times.com, and Yahoo's news section, provide message board and chat options for people to express opinions and debate political affairs. There is a wealth of public interest and social issue websites that provide access to email lists and message board discussion. There are even sites that exist solely to offer a public forum for political conversation, such as Quorum.org. Analysis of surveys in the United States suggest that during the political election year of 1996, 28% of Internet users participated in at least one of the following four activities: engaging in conversation online about the election, receiving email about a campaign, sending or receiving email to or from government officials, and sending email to others regarding a campaign (Katz and Rice 2001). A recent survey by the Democracy Online project at George Washington University (Democracy Online Project 2002) found that 43% of Internet users surveyed reported that in the past four years they went online to get information or discuss politics and government. The research that looks at political conversation online has begun to establish the content of political chat spaces, such as Usenet, email lists, message boards, and chat (Benson 1996; Davis 1999). We are only beginning to understand, however, which people are more likely to engage in such conversations online. We do not know if some people are drawn to certain communication channels online, such as listservs versus bulletin boards. We also do not know if the same people who engage in these online political conversations are those who engage in these types of conversations offline, and we do not know why they engage in such conversation online in the first place. The question that drives this study revolves around the larger query: Why do people talk politics online, given that face-to-face political conversational opportunities still exist? I take as my launching point the literature that suggests: (a) that many people find political conversation to be taboo in most social settings (Eliasoph 1998), (b) that many people who talk politics do so primarily with close friends or family members (Wyatt, Katz, and Kim 2000), and that (c) there are particular kinds of people who are more likely to talk politics than others (Scheufele 1999). Verba, Schlozman, and Brady (1995) argue that political discussion does not require a great deal of resources: "It demands neither money nor organisational skill nor even the capacity to make a public presentation" (p. 362). Yet, there are studies suggesting that people eschew political conversations, particularly when they are considered to be "public" conversations. Eliasoph's (1998) ethnographic work found that the members of volunteer and social clubs she studied felt a taboo around political conversation in group settings, particularly political subjects that could be considered structural and systemic. She observed that in individual, one-on-one conversations in "private," however, these members would engage each other in conversations about public affairs, current events, or social issues. She uses Goffman's (1959) theatre terms of "backstage" and "frontstage" to characterise the behaviour of people. When people perceived themselves to be frontstage, that is, in public, they communicated politics in a restrained matter, avoiding conflict and comments that implicated larger social structures—a finding similar to that found by Wyatt et al (1996) in their survey analysis of the main reasons people have for avoiding political expression. When people perceived themselves to be in private, they offered more unconstrained political discussion of social issues and public affairs. Thus, in public, where the discussants are more diverse, people avoid political topics. The vibrant public sphere, which requires such conversation for sustenance, is being deprived. If many people feel that political conversation is only appropriate in limited social settings, it becomes clear why Wyatt, Katz, and Kim (2000) found that most political conversation occurs in the home. In one study, they argue for the importance of casual conversation to political participation (Kim, Wyatt, and Katz 1999). In another study (Wyatt, Katz, and Kim 2000), they provide data that suggest an interrelationship between more traditionally conceived political conversation (foreign affairs, government, legislation, economy) with that of personal conversation (family matters, sports, entertainment, religion), and they measure where people report having such conversations. They conclude that the home is where most political and personal conversation occurs. This leads them to state that "paradoxically, home appears to be an integral part of the public sphere - the very point, in fact, where the public sphere and the family meet to form a life-world more integrated than Habermas (1962/1989) conceived" (p. 89). Home, the heart of the private sphere, allows us to participate in the public sphere with the aid of communication technology. People can choose to be public or private inside the private confines of the home. So, then, who is most likely to engage in political conversation? The researchers who have built and tested models that predict political conversation offer a set of demographic characteristics, a set of political characteristics, and a set of media use and communication characteristics that factor into such a model. Of the demographic characteristics, age, gender, income, and education are typically included. Scheufele (1999) hypothesised that higher education levels coupled with higher income would have positive effects on political conversation. Education imparts knowledge and cognitive skills necessary for understanding the political landscape (Converse 1988; Wolfinger and Rosenstone 1980). Scheufele also predicted that "younger men with higher levels of education and income are likely to express their opinions than other groups" (p. 29). Straits' (1991) model also included education and age. Similar to education, as one grows older, one is likely to attain life skills that enhance knowledge and cognitive ability. Straits, however, did not believe race or gender would have any significant effect. In studies of political participation, differences between men and women and between whites and non-whites have been negligible (Bennett and Bennett 1986). Straits also argued that one should take into account the "supply of appropriate discussants in the surrounding social environment" (p. 434). A lack of suitable discussion partners may restrict the amount of political conversation one engages in. Therefore, being married should have a positive affect on political discussion. When testing their predictions, Scheufele (1999) found that the demographic variables were not significant. Straits (1991) found that the background characteristics did have an effect on the amount of political talk when the model only included the background variables. When the variable of marriage was added, education and age dropped out of the equation as having a statistical effect. As predicted, people with spouses also were more likely to engage in more frequent political talk. Wyatt et al (1996) found that education, income and gender were related to freedom of expression, such that better educated, wealthier men reported feeling more comfortable speaking out. Interest in politics is an obvious predictor of whether one engages in political conversation (Verba, Schlozman, and Brady 1995). People who are interested in politics will more likely engage in political conversations. Knowledge about politics also should be positively related. Knowledge plays an important role in whether someone is willing to communicate about an issue. Prior political participation should also predict political conversation, as people who engage in political acts are likely to develop what Verba et al call "civic skills." Such development of these skills should contribute to both knowledge and willingness to talk about politics. Finally, political ideology and party strength should have some effect on political conversation. Political ideology, where one sits on the liberal/conservative dimension, may have some effect upon political talk. Straits (1991) found that liberals were more likely than conservatives to engage in political conversation. Party strength has been found as an important component in predicting political participation (Bennett and Bennett 1986). Finally, internal political efficacy is commonly viewed as a key variable in political conversation and political participation studies. The level of efficacy a person feels, the more they support the political system and feel they can participate in it (Gamson 1968). Communication and media variables would seem to be important aspects to include in a model because the dependent measure is a communication variable. It stands to reason that certain communication patterns influence others. Straits (1991) predicted and found that newspaper use was positively related to political discussion frequency, because "those who follow politics in newspapers should be motivated to share, exchange, and compare this information with others" (p. 434). Scheufele (1999) similarly included media use in his model, explaining that "exposing oneself and attending to public affairs or hard news provides respondents with political information necessary to engage in political talk at all" (p. 31). Political talk radio listening should also be strongly correlated with political conversation for the same reasons. Alongside media use, general interpersonal discussion might also provide information and indicate some predisposition to articulate ideas to others, and so should be a factor in political discussion. Given what we know so far about political conversation in face-to-face settings, some questions emerge in light of the vibrant community of political conversations occurring on the Internet. First, are there people who are talking politics online who are not doing it face-to-face or are the same people talking online who also do so face-to-face? That is, do people who talk politics online and people who talk face-to-face share the same characteristics? Why do people talk politics online? What do they offer as reasons for engaging in such a behaviour, given that political conversation is often seen as a taboo behaviour? I hypothesise that there are people who use the Internet as their medium of choice to engage in political conversations who avoid engaging in similar conversations in face-to-face contexts. I also hypothesise that the people who engage in political conversation online are categorically different than those who do so face-to-face. If the medium matters (Heidegger 1977; Williams 1975), then there should be differences in the people drawn to use the network-mediated context as their medium of choice for political conversation. I also posit that people are drawn to political talk online because they have an opportunity to hear diverse opinions. If most people who talk politics primarily do so with friends and family, those people are likely to share similar opinions. Online political talk gives one an opportunity to observe a broader range of opinions and arguments. A second reason people may give for engaging in online political talk is that it gives them an opportunity to express their opinions to diverse others. Online political talk gives one a "soapbox" to express opinions to an audience, without the same perceived social risks that expressing opinions face-to-face might bring. Methods and Measures In order to investigate these hypotheses, I conducted secondary analysis on surveys conducted with a group of people involved in a year-long study of political conversation and politics.1 Called the Electronic Dialogue Project and run at the Annenberg School for Communication at the University of Pennsylvania, it was designed to examine whether group deliberation produces deeper and more thoughtful public engagement with political issues. Sixty groups, drawn from a random sample of U.S. citizens both with and without Internet access, engaged in a year-long project during 2000 in which they engaged monthly in real-time electronic discussions, using Web TVs to gain access to the discussions, about issues facing the country and the unfolding presidential campaign. They answered surveys before and after each discussion. Two baseline surveys were conducted, the first from February 8 to March 1, and the second from March 10 to 23 of 2000. The surveys included measures of media use, interest in the presidential campaign, general political knowledge, political discussion, and a wide variety of political attitudes and opinions. Response rates to each of the baseline surveys were approximately 90 percent. All of the measures - except for party strength, political ideology, and political interest - used in my analysis come from the baseline surveys. The party strength, political ideology, and political interest questions were asked as part of the recruitment survey. These measures were used in the crosstabular and logistic regression analyses, and all were taken before the participants began the experimental portion of the study. In determining why people might talk politics online, I use one post-discussion survey of the experimental group administered in June 2000, which asked them to identify what they like and dislike most about the online political discussion they participated in. Overall, 51% of those recruited agreed to participate in the Edialogue project, completed the consent forms, and completed the baseline surveys. The final number of recruited project participants was 2014. Of these, 906 were assigned to a main discussion panel, in which they would meet online once a month over nine months in a pre-assigned discussion group of approximately 10 other assigned members, 139 to a survey-only control panel, and the remainder to the pre/post "set-aside" group. All participants were administered the baseline survey. Analysis of the groups across key demographic variables compared to a random-digital dial sample of the national population at the Annenberg Public Policy Center indicates that our sample is representative of the national population. The Electronic Dialogue Project (N=3967) had slightly more white members, were slightly better educated, and slightly younger. Because the sample agreed to participate in a year-long study about politics, they were more likely to be interested in politics. Dependent Measures. The dependent variables I am interested in are communication variables: asking people about political talk with friends and family, with acquaintances, and online. The political talk online question asked if the respondent had used the Web to go to a chat room, newsgroup, discussion forum, bulletin board, or similar interactive service to discuss political or social policy issues, current affairs, or political campaigns in the past 12 months. Twenty-one percent (N=338) of the total asked (N=1581) said they had gone online to talk politics in the past twelve months (M=.21; SD=.41). A comparison of the Edialogue sample with a Pew Internet and American Life (Pew Internet and American Life Project 2001) sample similarly suggests that roughly 20% of people in the U.S. with Internet access go online to "chat in a chat room or an online discussion space." The interpersonal political talk questions asked participants to provide initials of two close friends or family members with whom they discuss politics or public affairs and provide the number of days in a typical week they talk politics with them. Respondents were also asked to provide the initials of two acquaintances "people at work or others you see just going about your day" with whom they discuss politics or public affairs. As with the questions about friends and family discussion partners, the frequency of conversations with acquaintances was also asked. Of friend and family discussion partners, the number of partners and number of days was combined into a single dichotomous variable where "0" is no political talk and "1" is the reporting of talk with face-to-face discussion partners (M=.86; SD=.35); The same manipulation was done of the political talk with acquaintances question to create a dichotomous variable (M=.49; SD=.50). By making the interpersonal, face-to-face political talk variables dichotomous, I could more easily compare the results in the regression with the political talk online variable, which was asked of participants as a dichotomous variable. Political Interest. Interest is a combination of two questions: how closely someone reports following public affairs and how much someone cares who wins the 2000 presidential election with "1" being low and "4" being high. The two variables were averaged to create a scale of political interest (Cronbach alpha = .62; M=3.20; SD=.71). General Political Knowledge. The survey employed several different measures of political knowledge, including a general political and civics knowledge battery of 10 items, which were used for this analysis. Questions asked about the political parties, the political process of determining a law, the main duty of Congress, how the Senate and House overrides a presidential veto, and what positions Trent Lott, Al Gore, and William Rehnquist hold. All items were scored with "1" being correct and "0" being incorrect. The items were averaged to create a scale (Cronbach's alpha = .81; M=.71; SD=.27). Party Strength and Political Ideology. Party identification was measured by asking people to identify themselves with one of the two major political parties. This original variable was then recoded to create a political strength variable with value range from "0" as no strong party identification and "3" as very strong party identification (M=1.92; SD=1.01). Political ideology, similarly, is derived by asking respondents to identify themselves along the liberal/conservative dimension, with "1" being very conservative and "7" being very liberal (M=3.77; SD=1.60). Efficacy. Three questions measuring whether the individual has any say in what the government does, whether public officials care what the individual thinks, and whether politics is too complicated for the individual were coded so that "5" represents the efficacious response and "1" represents the low-efficacious response. The three items were added together to create a scale (Cronbach's alpha=.66; M=7.57; SD=2.53). Political Participation. Participation in political activities was measured through people's responses regarding whether they attended any political meetings, rallies, speeches, dinners or similar events in support of a particular candidate, did any other work for a candidate, gave money to a candidate, wore a candidate's campaign button, or put a campaign sticker on a car or placed a sign in a window with "1" as doing the act, and "0" as not doing the act. The items were averaged to create a scale (Cronbach's alpha = .61; Mean=.06; SD=.16). Media Variables. Participants were asked how many days in the past week they watched national network news (M=3.30; SD=2.50), watched local TV news (M=4.41; SD=2.45), watched cable news on TV (M=2.63; SD=2.58), read a daily newspaper (M=3.91; SD=2.72), or listened to political talk radio (M=1.67; SD=2.26). Items range from "0" days a week to "7" days a week. The three television news variables were combined into a single index of the mean of the number of days of television news exposure (M=3.43; SD=1.93). Participants were also asked about the number of hours they used email per week (M=3.32; SD=5.81), and how many hours they used the World Wide Web per week (M=6.33; SD=9.10). Participants also were asked how long they had been using the Web, with values of: "1" in the last two months, "2" within the last six months, "3" a year ago, "4" two or three years ago, and "5" more than three years ago (M=3.21; SD=1.50). Demographic Variables. Included in the demographic measures are age, income, education, marital status, number of children, gender, and race. The age range of participants was 18 to 92 (M=42). For the analysis age was recoded into 7 approximately equal categories (M=3.26; SD=1.54). Income was recoded into 6 categories of equal numbers of respondents where "1" is a low value and "6" is a high value ($0-$25,999, $26,000-$39,999, $40,000-$50,999, $51,000-$67,999, $68,000-$90,999, and $91,000 and higher) (M=3.45; SE = 1.73). Participants offered their years of education, which was recoded into 4 categories: 8 years or less, 9-12 years, 13-16 years, and more than 16 years. Marital status was asked of respondents and recoded into "1" married and "0" not married (M=.64; SD=.48). Respondents were asked to provide the number of children in their household, which was kept in a continuous variable (M=.35, SD=1). Our sample had equal numbers of men and women (M=.5; SD=.5). Race was recoded into a dichotomous variable of white ("1") and non-white ("0") (M=.8; SD=.4). Results The first question is whether the same people who talk politics online are also doing so in face-to-face contexts. As I suggested earlier, I believe there are people who are using the Internet to talk politics who do not do so in person. A cross-tabulation of political talk online and political talk with friends and family indicates that for this study there are people who talk online who report not having any discussion partners who are friends and family (see Table 1). Of those who went online in the past year to talk politics, 16% report having no political discussion partner who is a friend or family member (N=48). In looking at political talk online and political talk with acquaintances, there exists a larger number of people who engage in political conversation online but who do not have political conversations with acquaintances. Half of those who report talking online, report having no acquaintances with whom they talk politics (N=150). From this, it is evident that there are people who go online to talk politics who eschew face-to-face political conversations either with friends and family or with acquaintances, and a full 50% who talk politics online report no acquaintances with whom they talk politics. Table 1: Political Talk Online Compared with Political Talk with Friends and Family, Acquaintances, and Face-to-Face Online Talk Do Not Talk Per cent Talk Per cent Total Talk with Friends and Family 1017 87 259 84 1276 Talk with Acquaintances 589 51 157 51 746 Talk Face-to-Face 1031 88 262 85 1293 Total 1165 307 1472 This conclusion, however, does not necessarily mean that there are people who engage in no interpersonal, face-to-face political conversation, but do engage in political talk online. It is possible that the people who talk with friends and family do not talk with acquaintances and vice versa. To determine if that is the case, another crosstabulation was done that indicates the number of people who talk face-to-face (combining the two variables of friends and family and acquaintance political talk) with those who talk online (see Table 1). This crosstabulation suggests that 12% of this sample indicate they have no political discussion, online or off, and 85% indicate they have political conversation both online and off. Of particular interest for this analysis, 15% indicate that they had no face-to-face discussion partners, but did go online to discuss politics. This finding gives some credence to the notion that there are people using the Internet for political conversation that avoid it in face-to-face situations. The second issue is whether those who talk online share the same characteristics as those who talk face-to-face. This is important to investigate because the people who talk online who are not doing so face-to-face may simply be people who talk face-to-face, but now use the Internet as their outlet for political conversation instead. The logistic regressions indicate, however, that there are differences between those who talk with friends and family, with acquaintances, and online (see Table 2). Three logistic regression models were created, predicting friends and family political talk, acquaintance political talk, and online political talk.2 The independent variables were the same set for each model (identified in the last section), in order to compare the models.3 The logistic regression model predicting political talk with friends and family (Chi-Square=205.34; Pseudo R2=.27) suggested that positive predictors included gender (female), political interest, general political and civics knowledge, and political talk with acquaintances. Age was negatively predicted. The variables of education, race, internal political efficacy, and media use and exposure variables were non-significant predictors. The logistic regression model predicting political talk with acquaintances (Chi-Square=180.01; Pseudo R2=.17) indicated that gender (male) and political talk with friends and family were positive predictors. There were no variables that negatively related to the dependent variable. The variables of age, education, race, internal political efficacy, political interest, general political and civics knowledge, and all media use and exposure variables were non-significant. The third model, political talk online (Chi-Square=155.65; Pseudo R2=.17) sug Table 2. Logistic Regression of Political Talk Variables with Independent Variables Predictors Talk with Friends and Family Talk with Acquaintances Talk Online Demographic Variables Age -.15* -.03 -.37*** Education .00 .11 -.09 Gender (male) -.66*** .33** .37* Race (white) .41 + .01 -.29 Political Variables Political Interest .61*** .21* .53*** Political Knowledge 1.45*** .48+ -1 .50*** Efficacy -.01 .04+ -.07** Communication Variables Talk Radio .01 .00 11*** F2F Friends/Family Talk ---- 2.26*** .01 F2F Acquaintance Talk 2.29*** ---- .13 Online Talk -.05 .11 ---- Web Use (hours) -.01 -.01 .04*** Intercept -.89 -3.73 -.63 Chi-Square 205.34 180.01 155.65 Pseudo R2 (in percent) .27 .17 .17 Note. N=1338; * p < .05, ** p <_.01, *** p < .001, + p < .10 gested that gender (male), political interest, political talk radio listening, and Web use were positive predictors. Age, general civics and political knowledge, and internal political efficacy were negative predictors. The non-significant variables were education, race, political talk with friends and family and political talk with acquaintances. In summation, the differences suggested in this model are these: people who talk with friends and family tend to be younger women who are politically interested and knowledgeable; people who talk with acquaintances tend to be men; and, people who talk online tend to be younger men who are politically interested but who are less knowledgeable about civics and general politics and are less efficacious. They tend to listen to political talk radio. Other variables that commonly appear in the literature, such as education, marital status, race, number of children, political ideology, and political participation were not significant for any of the dependent measures, and so were dropped from the regression model presented in Table 2.4 What is important to note is that although talking with friends and family strongly predicts whether someone talks with an acquaintance and vice versa, there is no relationship between in person and online talk; this further £ reinforces the idea that there are people who talk with acquaintances or friends and family that are different than those who do so online. The models offered by previous scholars suggest men engage in more political rn talk than women (Huckfeldt and Sprague 1995). The findings for this study sug- A 3 a gest the opposite, at least for political talk with friends and family. As expected, political interest and political knowledge both factor into whether someone is more likely to talk politics with friends and family; however, it does not factor into whether someone talks with acquaintances. It would appear that neither interest nor knowledge plays key roles in whether people engage in conversations with people who are co-workers, neighbours, or people at church. Although newspaper use was a significant predictor when all the variables were entered into the model, the question about newspaper use could have been interpreted as either newspaper reading online or offline, but was not specific. As a result, it is possible that people who read the newspaper online might not have reported it because of the way the question was worded, and so that variable was dropped from subsequent analysis. In looking at the online political talk regression, I found that gender was a key predictor—that men were more likely to be the ones who went online to post a comment on a political or civics matter. This is in keeping with findings of studies of Usenet in which it appears that most discussants are men (Davis 1999; Hill and Hughes 1998). Younger people are more likely to talk politics online, which is not surprising given that younger people are more likely to be online (Pew Internet and American Life Project Sept 21, 2000). Of the political variables, people who talk politics online may be less knowledgeable and efficacious than those who engage in face-to-face conversations, which was a surprising finding. On one hand, this may mean that there exist online less informed conversations from people who feel that they cannot make a difference in the political structure. On the other hand, this may mean that there are people who would normally not feel comfortable expressing their opinion in interpersonal situations, for fear that they would be challenged or appear unknowledgeable on a subject, who feel free to do so online. Although uninformed opinion expression is often viewed as highly problematic for healthy deliberation, it might also mean that there are people who are expressing their opinions who would be disinclined to do so otherwise. Similarly, efficacy, particularly as it relates to trust, has been viewed as a powerful motivator for people to engage in political activities (Gamson 1968). Yet it may be that people who feel less efficacious have much to complain about and the Internet provides a forum through which they can voice their frustrations with the current political scene. Finally, why do people talk politics online? What do they offer as reasons for engaging in such behaviour, given that political conversation is often seen as taboo? I posit that people are drawn to political talk online because they have an opportunity to hear diverse opinions. If political talk partners are primarily friends and family, there might not be much diversity of opinion. Online political talk gives one an opportunity to observe a broader range of opinions and arguments. A second reason people may have for engaging in online political talk is that it gives them an opportunity to express their opinions to an audience of other people. Online political talk gives one a "soapbox" to express opinions to others who may be quite different from each other. To confirm or reject this, I turn to the online discussions themselves and those who participated in them. Participants in the post-event survey in June, 2000 were asked in an open-ended format what they liked and disliked about the political discussions they were participating in through the experiment (see Table 3).5 The most common response participants gave was hearing or learning about others' opinions, accounting for 29% of the total number of positive comments made. The second most common response, comprising 12% of the positive comments, was being able to express ones' own opinion. Another common response people gave was enjoying hearing from people who are in different parts of the country, comprising 10% of the positive comments. People's comments of what they liked most about the discussion included statements such as these: • "Talking with other people about issues that effect us all." • "Getting other people's ideas from other parts of the country." • "An overall viewpoint from all the participants. You get a chance to agree or disagree." • "Speaking my opinion." • "Participating in a discussion with people who I don't know and who don't know me. I feel free to say what I really feel without any fear of criticism or reservation. My feeling is it doesn't matter what they say about my opinion because "they" are words that appear on the screen, although I do consider what is said." Respondents' comments suggest that they enjoyed having the opportunity to talk with people they did not know and would probably never get a chance to interact with face-to-face. They seemed to enjoy a conversation devoted entirely to political issues, in which it is the norm to express a point-of-view and to agree or disagree with others' points-of-view. The person who provided the last quotation indicated a freedom from a concern that I believe troubles people in face-to-face political discussions—fear of criticism and the social isolation that may come from that. Even though there is disagreement in the online conversations, the normal social repercussions that might happen in a face-to-face encounter do not have the same magnitude, if they exist at all, online. Table 3: Open-Ended Response Categories of Liked Most/Disliked Most About Online Discussion "Liked Most" Number of "Disliked Most" Number of Comments Comments Hearing Others' Opinions 128 Fast Pace/Keeping up with Comments 62 Expressing Own Opinions 53 Typing Skills 45 Different Parts of Country 45 Too Short 25 General Interaction 45 Too Slow 24 People/Atmosphere 42 Technical Problems 23 Agreement/Disagreement 31 Lack of or Too Much Focus 18 Subject/Content 21 Lack of Depth 16 Technology 15 Group Size 10 Making a Difference 11 Not Seeing Results 6 Miscellaneous 52 Too Much Agreement (Boring) 5 Total Comments 443 Time Commitment 5 Total Participants 309 Too Long 3 Moderator 3 Lack of Knowledge of Participants 3 Misc 111 Total comments 359 Total Participants 299 People also had negative comments about the discussions, although they provided fewer of these. The negative comments were more difficult to categorise, giving the "miscellaneous" category the most responses. The second most common response, accounting for 17% of the negative comments, related to the nature of the synchronous online conversations and the pace at which the conversation "scrolled." Rare were criticisms about the experience of expressing an opinion or hearing the opinions of others. These open-ended comments, although specifically directed to the experimentally crafted and highly structured discussions, still provide some indication of why people might enjoy political conversation online. The artificial nature of the experiment does not hinder the interpretation that people who choose to go to a political discussion space online also experience similar thoughts about what they like and dislike. Discussion I believe that there are at least three interrelated characteristics of the online conversation context that serve to draw people into using the Internet as a medium for political conversation: an absence of non-verbal cues, which leads to a lowered sense of social presence, and a heightened sense of anonymity. By "Internet," I refer in shorthand to the communication channels available, such as email, message boards, newsgroups, and real-time "chat." The channel characteristics I refer to hold for those channels that offer interactive, human-to-human communication. The Internet characteristics of reduced cues, reduced presence, and heightened anonymity may mitigate perceived social risks that hinder some people from engaging in political conversations in face-to-face settings, particularly amongst those who are acquaintances or strangers. Those who have a desire to talk politics with others can do so online without the risks or fears that may prevent them from doing so in face-to-face settings. That the social interaction context online has reduced social cues was one of the observations noted by the pioneers of computer-mediated communication research. Communicators cannot see their fellow conversants in most CMC exchanges, so cues of skin colour, body posture, vocal quality, tone, and other non-verbal cues we use to make judgements about those with whom we engage in conversation are largely absent. McKenna and Bargh (2000) explain: "when an individual posts an article in a newsgroup or enters a chat room full of strangers, he or she may well feel that his or her actions will be submerged in the hundreds (or thousands) of other actions taking place there" (p. 60). Although the notion that there are no social cues online is a contested claim, the communicative information we garner through proxemics, vocalics, kinesics, and facial expressions is missing in textual interactions that occur over email, synchronous chat, or on message boards. This, it can be hypothesised, alters the communication context and communication experience. One of the ways the communication experience may be altered is in a reduced sense of social presence. There are a range of notions of what social presence is. I utilise the notion that social presence is the degree to which media give audiences the sense that they are present in the narrative or action depicted in the medium (Lombard and Ditton 1997). According to the classic study by Short, Williams, and Christie (1976), the capacity of a medium to transmit non-verbal cues contributes to the perception or attitude of the user to the medium. If the social context in a textual discussion space strips non-verbal communication cues, this may entail that people experience emotional and cognitive distance in their social interaction as compared to similar face-to-face interactions. The effects of such reduced presence may include people feeling they have greater control over the social interaction. A third characteristic of the social context online may be that people who engage in discussion online remain physically in the private sphere, i.e. at home or at work. Their bodies remain in their computer chairs, facing a monitor and keyboard, in a space that is often private—the home or the semi-public domain of a workplace. The physical body is sheltered from the visual engagement of others, and is comforted by the items that indicate the private—hot tea, a bar of chocolate, photographs of family, a comfortable chair. Because a person engaging in online conversation is veiled behind a computer screen and in a private setting, they may feel a heightened level of anonymity and comfortability they would not feel in a physically public setting talking about political subjects. As a result, people may feel more in control, bolder, safer engaging in conversation online. These three characteristics in conjunction or separately may liberate people from their concerns to engage another in political conversation online. The Internet conversation context seems to create distance between interlocutors—distance that may liberate some people to express views and ideas that they would not do face-to-face because of the perceived risk of social repercussions. It should be noted that this is speculation, and these proposed characteristics need further empirical testing to determine if Internet users experience them as such, and if the characteristics have these effects. A problem with some CMC research, that I want to avoid, is the assumption that technology brings forth new social arrangements and social effects; that is, many researchers of the Internet explicitly or implicitly espouse a view that is technologically deterministic. Williams (1975) and Heidegger (1977) make compelling arguments for why technological determinism is problematic. Heidegger defines technology both as a means to an end and as a human activity. Humans bring technology into being for certain reasons. Those reasons may change, and the ways technologies are used may not resemble their original intent, yet technology should not be understood as existing outside of and before human needs and human actions. Similarly, Williams argues that mass communication technology emerged out of growing necessity as individual mobility increased and organisations increased in scale. New communication technologies were needed in order to meet these changing social conditions. Technology is developed through a reiterative process of social needs creating demand for new technologies, which are in turn developed and used, in part to meet those social needs and in part to create new social needs. The Internet, whether viewed as a mass communication or a social interaction medium (or both), continues the trajectory of satisfying needs. The same cultural forces that Williams identified still exist: expansion of organisations, mobility and privatisation of individuals. Out of these cultural forces emerged the Internet. It, too, will meet some needs and create new ones. Some people will be drawn to use it in certain ways, others may not. With regard to political conversation, I believe that it is vitally important for democratic government that people engage in political conversations with acquaint- ances and strangers. If people only talk politics with close friends and family members, then participatory democratic theory is unfulfilled. Participatory democratic theory holds that people are highly interconnected, and that social relations enable people to share and diverge on interests and to work towards social action. The key ingredients to participatory democratic theory are those social relations that bind people together (Dewey 1946; Rucinski 1991). Cooley (1909) explains "in politics communication makes possible public opinion, which, when organised, is democracy" (p. 85). Participatory democratic theory rests on an assumption of the generation of public opinion and, by extension, a public sphere. A healthy public sphere seems to require both a multitude of publics (Calhoun 1997; Fraser 1993) and a free exchange across those various publics (Dewey 1946). That free exchange occurs through weak ties, in which people who are not close friends, but rather acquaintances, share information, express opinions, and continue the conversation. More importantly for public sphere theory itself, the Internet appears to provide a place for people to talk who are simultaneously public and private, which does not sit squarely with traditional notions of the public. In the writings of Arendt (1958) and of Habermas (1962/1989) both place and action constitute the public sphere. The conversations that helped spawn the public sphere occurred in publicly accessible spaces. Habermas suggested that the bourgeoisie of late 17th and early 18th Century Europe consumed news and culture, deliberated news and culture, and even moved to political action in the "public" spaces of coffeehouses and salons. Even earlier in the 16th Century, Zaret (2000) contends that taverns, inns, and alehouses were common spaces for the oral transmission of news and opinion on that news, because of their accessibility to people, primarily men, who came from different social stations and occupations. Yet, place alone does not constitute the public or private spheres. The kinds of activity that occur in a space also signify "public" or "private." The action of deliberation or political discussion for its own sake that occurs in spaces accessible to a heterogeneous group, i.e. a public place, constitutes the public sphere. Similarly, the actions of eating or having sex at home is a private sphere activity. Presumably, a researcher would observe and label talking on a cell phone and walking on a sidewalk as a private-sphere activity, while talking politics on the Internet in a chat space when sitting at a computer at home becomes a public-sphere activity. The kind of activity and the kind of space the activity occurs in matter to whether it is considered a public or private sphere activity. There are people who engage the Internet as a communications vehicle through which they express political points of view and discuss them with geographically diverse others while never leaving the confines of the home or other typically private place. The simultaneous public and private nature the Internet conversation context provides may enable people to talk politics, people who for whatever reason, do not typically feel comfortable doing so in a public, face-to-face setting with acquaintances (Eliasoph 1998). The social context which now includes the Internet, thus, enables a group of people to participate in political conversation who may not have had an outlet or who may not have been comfortable doing so in interpersonal settings. A potentially new group of political conversationalists may be emerging with the help of the online context, thus adding voices to the public sphere, albeit ones in which those voices' bodies remain physically "private." Conclusion What I have found through this analysis is that a group of people have gone online to talk politics who do not report having face-to-face discussion partners, and half of those who report talking politics on the Internet report having no acquaintances with whom they talk politics. It seems that people choose the Internet as an avenue for talking politics with strangers and acquaintances who would avoid doing so in similar social face-to-face settings. The study also suggests that people talk online because they enjoy hearing what diverse others have to say. They appreciate being able to express their own points of view and they enjoy hearing others' views. This study provides some encouraging news about the relationship that may exist between the Internet and the political process in a quickly growing field of studies that suggest the Internet will have either a negligible (Margolis and Resnick 2000; Resnick 1998; Scheufele 2001) or a negative effect (Brook and Boal 1995; Doheny-Farina 1996; White 1997) on political participation and civic engagement. One problem with this study is the imprecise nature of the measure of online political conversation. Asking people if they recall talking politics in the past year online is quite a different question than asking people if they had a political conversation with friends and family or acquaintances in the past week. The finding that there are people who talk online who do not do so face-to-face may be a result of the fact that a person said they did not talk last week, but perhaps did so in the past month. The question, however, specifically directs people to think of two friend and family members and two acquaintances with whom they talk politics. Over 200 people provided no people with whom they talk politics with, giving some indication that the question is tapping larger political conversation patterns. Work still needs to be done to determine why people go online of their own volition to talk politics. This study investigated some motives of online political talk derived from an experiment in which people were asked to discuss politics in a small group setting. More study is needed to determine what larger set of motives are satisfied by online political conversation, and such study should be conducted on actual Internet users who engage in political discussion in publicly accessible online spaces. Understanding who talks and why they talk politics online is important because such behaviour may have social implications on future political behaviour. The Internet may provide a new context for political conversation for those who would not normally engage in face-to-face political conversations. Moreover, the Internet offers the technological marvel of at least partially erasing physical and temporal boundaries that could prevent people from easily conversing with each other. People from disparate geographic, temporal, and ideological situations can converge online to exchange points-of-view. Because of these factors, the Internet may offer a new space for public discussion where voices can enter the public sphere that would otherwise remain silent. Notes: 1. Data for this study came from the Edialogue Project at the Annenberg School for Communication, University of Pennsylvania, Vincent Price and Joseph N. Cappella, Principle Investigators. The Edialogue Project was funded in part by a grant from the Pew Charitable Oì m 3. All the media variables except political talk radio and Web use were non-significant in any of the logistic regression models and were removed from further analysis. 4. What is presented in Table 2 are only the key variables that have some significance in the model. When all the original variables are added into the model, race is significant for talk with friends and family. Whites are more likely to engage in face-to-face talk than non-whites. However, the number of people accounted for in the model drops from 1,338 people in the shortened model to only 875 in the model with all variables; the percentage variance explained by the models was the same in both models. 5. Respondents' statements were broken into up to 3 distinct ideas. Ideas were then coded into the categories, which were generated deductively. Ten categories were created for "likes" and fifteen for "dislikes" by three coders ("Likes" Cronbach Alpha = .83; "Dislikes" Cronbach Alpha = .74). References: Arendt, Hannah. 1958. The Human Condition. Chicago, IL: University of Chicago Press. Bennett, Stephen E. and Linda Bennett. 1986. Political Participation. In S. Long (ed.), Annual Review of Political Science, 157-204. Norwood, NJ: Ablex. Benson, Thomas W. 1996. Rhetoric, Civility, and Community: Political Debate on Computer Bulletin Boards. Communication Quarterly 44, 359-378. Brook, James and Iain A. Boal. 1995. Resisting the Virtual Life. San Francisco, CA: City Lights Books. Calhoun, Craig C. 1997. Introduction: Habermas and the Public Sphere. In C. C. Calhoun (ed.), Habermas and the Public Sphere, 1-50. Cambridge, MA: Massachusetts Institute of Technology Press. Converse, Philip E. 1988. Perspectives on the Democratic Process. Michigan Quarterly Review 27, 285-299. Davis, Richard. 1999. The Web of Politics: The Internet's Impact on the American Political System. New York, NY: Oxford University Press. Davis, Richard and Diana Owen. 1998. New Media and American Politics. New York, NY: Oxford University Press. Democracy Online Project. 2002. Online Politically Active Citizens Not Worried About Losing Privacy on the Internet [WWW]. George Washington University, December 6, 2001 2002 [cited April 15 2002]. Available from http://democracyonline.org/mediacenter/ dec62001 release.shtml. Dewey, John. 1946. The Public and Its Problems: An Essay on Political Inquiry. Chicago, IL: Gateway Books. Doheny-Farina, Stephen. 1996. The Wired Neighborhood. New Haven, CT: Yale University Press. Eliasoph, Nina. 1998. Avoiding Politics: How Americans Produce Apathy in Everyday Life Cambridge: Cambridge University Press. Fishkin, James S. 1995. The Voice of the People: Public Opinion and Democracy New Haven, CT: Yale University Press. Fraser, Nancy. 1993. Rethinking the Public Sphere: A Contribution to the Critique of Actually Existing Democracy. In B. Robbins (ed.), The Phantom Public Sphere, 1-32. Minneapolis, MN: University of Minnesota Press. Gamson, William A. 1968. Power and Discontent. Homewood, IL: Dorsey Press. Goffman, Erving. 1959. The Presentation of Self in Everyday Life. Garden City, NY: Doubleday. Habermas, Jurgen. 1962/1989. The Structural Transformation of the Public Sphere: An Inquiry into Trusts. The author would like to thank Vincent Price and Nicholas Jankowski for guidance on earlier drafts. 2. The online political talk variable was asked as a dichotomous variable in the survey. The face-to-face talk variables were originally asked as ordinal, but were converted into a dichotomous variable in order to better compare the three models. a Category of Bourgeois Society. Cambridge, MA: Massachusetts Institute of Technology Press. Hart, Roderick P 1987. The Sound of Leadership: Presidential Communication in the Modern Age. Chicago, IL: University of Chicago Press. Heidegger, Martin. 1977. The Question Concerning Technology and Other Essays. New York, NY: Harper & Row. Hill, Kevin A. and John E. Hughes. 1998. Cyberpolitics: Citizen Activism in the Age of the Internet. Lanham, MD: Rowman & Littlefield Publishers, Co. Huckfeldt, Robert and John Sprague. 1995. Citizens, Politics, and Social Communication. In R. S. J. Wyer (ed.), Cambridge Studies in Political Psychology and Public Opinion. Cambridge: Cambridge University Press. Jamieson, Kathleen Hall. 1988. Eloquence in an Electronic Age: The Transformation of Political Speechmaking. New York, NY: Oxford University Press. Jamieson, Kathleen Hall. 1996. Packaging the Presidency: A History and Criticism of Presidential Campaign Advertising (3rd ed.). New York, NY: Oxford University Press. Jankowski, Nicholas and Martine van Selm. 2000. The Promise and Practice of Public Debate in Cyberspace. In J. van Dijk (ed.), Digital Democracy: Issues of Theory and Practice, 149-165. Thousand Oaks, CA: Sage. Katz, Elihu and Paul F. Lazersfeld. 1960. Personal Influence: The Part Played by People in the Flow of Mass Communications. New York, NY: Free Press. Katz, James E. and Ronald E. Rice. 2001. The Internet, 1995-2000. American Behavioral Scientist 45, 405-419. Kim, Joohan, Robert O. Wyatt and Elihu Katz. 1999. News, Talk, Opinion, Participation. Political Communication 16, 361-385. Lombard, Matthew and Theresa Ditton. 1997. At the Heart of It All: The Concept of Presence. Journal of Computer-Mediated Communication 3(2), online document. Margolis, Michael and David Resnick. 2000. Politics as Usual. In B. Sinclair (ed.), Contemporary American Politics. Thousand Oaks, CA: Sage. Mckenna, Katelyn Y. A. and John A. Bargh. 2000. Plan 9 from Cyberspace. Personality and Social Psychology Review 4, 57-75. Mutz, Diana C. 1998. Impersonal Influence: How Perceptions of Mass Collectives Affect Political Attitudes. Cambridge: Cambridge University Press. Page, Benjamin I. 1996. Who Deliberates? Chicago: University of Chicago Press. Pew Internet and American Life Project. 2002. Internet Activities [WWW]. Pew Internet and American Life Project, 2001 [cited April 9 2002]. Available from http://www.pewinternet.org/ reports/chart.asp?img = Internet_Activities.jpg. Pew Internet and American Life Project. Sept 21, 2000. Who's Not Online: 57% of Those without Internet Access Say They Do Not Plan to Log On. Washington, DC: Pew Research Center for the People and the Press. Resnick, David. 1998. Politics on the Internet: The Normalization of Cyberspace. In T. W. Luke (ed.), The Politics of Cyberspace, 48-68. New York, NY: Routledge. Rucinski, Dianne. 1991. The Centrality of Reciprocity to Communication and Democracy. Critical Studies in Mass Communication 8, 184-194. Scheufele, Dietram A. 1999. Deliberation or Dispute? An Exploratory Study Examining Dimensions of Public Opinion Expression. International Journal of Public Opinion Research 11 , 25-58 Scheufele, Dietram A. and Nisbet, Matt. 2001. Being a Citizen Online: New Opportunities and Dead Ends. Paper read at American Political Science Association, at San Francisco, CA. Short, John, Ederyn Williams and Bruce Christie. 1976. The Social Psychology of Telecommunications. New York, NY: Wiley. Straits, Bruce C. 1991. Bringing Strong Ties Back in Interpersonal Gateways to Political Information and Influence. Public Opinion Quarterly 55, 432-448. Streck, John M. 1998. Pulling the Plug on Electronic Town Meetings: Participatory Democracy and the Reality of Usenet. In T. W. Luke (ed.), The Politics of Cyberspace, 18-47. New York: Routledge. Verba, Sidney, Kay Lehman Schlozman and Henry E. Brady. 1995. Voice and Equality. Cambridge, MA: Harvard University Press. White, Charles S. 1997. Citizen Participation and the Internet: Prospects for Civic Deliberation in the Information Age. The Social Studies 88, 23-28. Williams, Raymond. 1975. Television: Technology and Cultural Form. New York, NY: Schocken Books. Wolfinger, Raymond E. and Steven J. Rosenstone. 1980. Who Votes? New Haven, CT: Yale University Press. Wyatt, Robert O., Elihu Katz and Joohan Kim. 2000. Bridging the Spheres: Political and Personal Conversation in Public and Private Spaces. Journal of Communication 50, 71-92. Wyatt, Robert O., Elihu Katz, Hanna Levinsohn and Majid Al-Haj. 1996. The Dimensions of Expression Inhibition: Perceptions of Obstacles to Free Speech in Three Cultures. International Journal of Public Opinion Research 8, 229-247. Zaret, David. 2000. Origins of Democratic Culture: Printing, Petitions, and the Public Sphere in Early-Modern England. Princeton, NJ: Princeton University Press.