Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During...

23
This article was downloaded by: [Washington State University Libraries ] On: 11 October 2014, At: 16:35 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Political Communication Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/upcp20 Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000 Joseph N. Cappella , Vincent Price & Lilach Nir Published online: 10 Nov 2010. To cite this article: Joseph N. Cappella , Vincent Price & Lilach Nir (2002) Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000, Political Communication, 19:1, 73-93, DOI: 10.1080/105846002317246498 To link to this article: http://dx.doi.org/10.1080/105846002317246498 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form

Transcript of Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During...

Page 1: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

This article was downloaded by: [Washington State University Libraries ]On: 11 October 2014, At: 16:35Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T3JH, UK

Political CommunicationPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/upcp20

Argument Repertoire as aReliable and Valid Measure ofOpinion Quality: ElectronicDialogue During Campaign2000Joseph N. Cappella , Vincent Price & Lilach NirPublished online: 10 Nov 2010.

To cite this article: Joseph N. Cappella , Vincent Price & Lilach Nir (2002)Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: ElectronicDialogue During Campaign 2000, Political Communication, 19:1, 73-93, DOI:10.1080/105846002317246498

To link to this article: http://dx.doi.org/10.1080/105846002317246498

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views ofthe authors, and are not the views of or endorsed by Taylor & Francis.The accuracy of the Content should not be relied upon and should beindependently verified with primary sources of information. Taylor andFrancis shall not be liable for any losses, actions, claims, proceedings,demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, inrelation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in any form

Page 2: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

to anyone is expressly forbidden. Terms & Conditions of access and use canbe found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 3: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

73

Political Communication, 19:73–93, 2002Copyright ã 2002 Taylor & Francis1058-4609 /02 $12.00 + .00

Argument Repertoire as a Reliableand Valid Measure of Opinion Quality:

Electronic Dialogue During Campaign 2000

JOSEPH N. CAPPELLA, VINCENT PRICE, and LILACH NIR

A new measure of opinion quality that we name “argument repertoire” (AR) is intro-duced and evaluated. AR refers to the relevant reasons that one has for one’s ownopinions and the relevant reasons that others with opposite opinions might have. Themeasure is shown to be reliable and to have construct validity. Those with elevatedAR also were more likely to attend on-line deliberative groups during the presiden-tial election and to contribute to those conversations. Those who participated in on-line deliberations tended to have higher AR scores on particular issues that werediscussed. The role of AR in deliberative political groups is explored.

Keywords campaign, deliberation, electronic discussion, opinion quality

This article is concerned with the development of a valid and reliable measure of opin-ion quality. Studies of opinion quality have taken two paths. The first focuses on inter-ventions that have the potential to improve the quality of opinions held by members ofthe public. These have included deliberative polls (Fishkin, 1991, 1995), issue forums,information-oriented polling, and civic journalism (Rosen, 1991), among others. Thesecond focuses on methods of measuring existing opinion that in principle rank opinionfrom higher to lower on some aspect of quality. These techniques have included filter-ing, consistency, coherence, and alternative forms such as consideredness. The researchwe report here focuses on the latter class of methods. We introduce a new and particularmethod for measuring opinion quality that we call argument repertoire (see also Lustic& Miodownik, 2000). The reliability and validity of the argument repertoire measure areevaluated in the context of a discussion-based intervention during the 2000 presidentialelection that we have dubbed Electronic Dialogue 2000.

Joseph N. Cappella is Professor, Vincent Price is Associate Professor, and Lilach Nir is adoctoral candidate, all in the Annenberg School for Communication at the University of Pennsyl-vania. This research is supported by grants to the authors from The Pew Charitable Trusts and theAnnenberg Public Policy Center of the University of Pennsylvania. The authors thank Yariv Tsfati,Jenny Stromer-Galley, Danna Goldthwaite, Tresa Undem, Son-Ho Kim, Clarissa David, EmilyWest, Masaki Hidaka, Anca Romantan, Eun-Kyung Na, Roselyn Lee, and Lisa Rand for theirassistance. Views expressed are those of the authors alone and do not necessarily reflect opinionsof the sponsoring agencies.

Address correspondence to Joseph N. Cappella, Annenberg School for Communication, Uni-versity of Pennsylvania, 3620 Walnut Street, Philadelphia, PA 19104-6220, USA. E-mail:[email protected]

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 4: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

74 Joseph N. Cappella et al.

Improving the Quality of Public Opinion

Democratic theory assumes that the public responsible for decisions about its electedleaders is well informed (Delli Carpini & Keeter, 1996; Habermas, 1981/1984). If col-lective decisions are not well informed, then these decisions are based on other consid-erations, perhaps ones that weigh less rational factors prominently. The implication, ifnot the reality, is that uninformed decisions will be inefficacious, at least, and harmful,at worst, to the public at large. Comprehensive studies of the public’s knowledge ofpolitics and social affairs suggest that the level of political knowledge is inadequate(Delli Carpini & Keeter, 1996) and that the quality of opinion is unorganized, incoher-ent, and uninformed (Bishop, Oldendick, Tuchfarber, & Bennett, 1980; Neuman, 1986).

In response, a number of scholars and activists have sought to improve the qualityof public opinion by providing information to those being polled either as a part of thepolling process itself (Kay et al., 1994; Neijens, 1987; Neijens, de Ridder, & Saris,1992) or as a part of a larger process of providing extensive information and the oppor-tunity to deliberate about that information (Fishkin, 1991, 1995). The fundamental aimof both approaches is to determine what public opinion would be in a full (or nearlyfull) information environment. This environment includes not only information gener-ally not possessed by the participants but also the opportunity to deliberate about, evalu-ate, and question this information in the presence of others with equal access to thesame informational resources.

High-quality opinions emerge through free and equal exchange, inviting argumentsfor all sides and granting to reasoned argument the power to shape collective choices(Arendt, 1958; Habermas, 1981/1984, 1962/1989). Discussion is seen as vital becauseit forces more careful consideration by challenging points of view—hence, those whodeliberate form better reasoned opinions. Moreover, discussion expands the repertoire ofconsiderations and arguments, and thus it fosters understanding of multiple points ofview (Gutmann & Thompson, 1996).

What is missing from these important and provocative innovations is any clear theoryabout what constitutes an effective intervention. One goal of deliberative polls is toimprove the quality of participants’ opinions. What constitutes a higher quality opinionis subject to some debate. Price and Neijens (1997, 1998) proposed a series of possiblecriteria that varied by outcome (e.g., stability) and process (e.g., extensiveness of infor-mation search) and by level (individual and collective). We propose “argument reper-toire” as one measure of opinion quality that is an outcome-oriented, individual-levelcriterion especially relevant to deliberative polling interventions.

Assessing Opinions for Their “Quality”

A number of researchers have criticized standard polling techniques for producing re-sponses that are ill-informed pseudo-opinions (Price & Neijens, 1997). Much mass opinionis thought to be disorganized (Converse, 1964) and without a basis in belief or knowl-edge (Herbst, 1993; Neuman, 1986; Zaller, 1992). A variety of methodological tech-niques have been proposed to minimize the influence of such unanchored opinions.They include the use of question filters, probes, “don’t know” alternatives, intensityratings, and other techniques (Converse & Presser, 1986; Schuman & Presser, 1981).These procedures are aimed at minimizing the overselection of weakly formed opinions,generated off “the top of the head” by respondents.

An alternative to eliminating weakly formed opinions from the response base is to

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 5: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 75

measure the degree to which opinion statements are anchored in beliefs—both support-ive and unsupportive of the focal attitude. This view assumes that opinions are closelyrelated to attitudes as affective responses and that these attitudes, in turn, are connectedin a cognitive network to other attitudes and concepts through beliefs (Cappella & Folger,1980; Fishbein & Ajzen, 1975; Pratkanis, 1989).

Fishbein and Ajzen (1975) suggest the use of open-ended elicitation procedures toextract salient beliefs (usually in the form of positive and negative consequences) thatare the basis of attitudes toward behaviors. Wyatt, Katz, and Kim (2000) have used“consideredness” as a measure of opinion quality based on judgments by interviewers ofthe reasons provided by interviewees for their opinions. Of course, the NES surveyshave included open questions about the positive and negative “considerations” that re-spondents have about political candidates (Zaller, 1992).

Argument repertoire is a measure of opinion quality that functions much like con-siderations and beliefs. It provides a way to rank stated opinions in terms of how wellanchored they are in reasons for one’s own opinion and for opinions that are opposed toone’s own.

Argument Repertoire as Opinion Quality

Our measure of argument repertoire is derived from the work of Deanna Kuhn (1991)on the nature of reasoning in daily life. She focuses less on what people think than whythey think what they do. In a detailed conceptual and empirical analysis, Kuhn main-tains the following: (a) argumentation is nothing more than reasoning to show the truthor falsity of some claim, (b) argumentation does not need to be dialogic to be studiedand evaluated, and (c) expressed argumentation is indicative of the structure of knowl-edge (i.e., attitudes and beliefs).

Kuhn carries out a deep and broad description of the nature of naive causal argu-mentation including the kinds and qualities of evidence in support of one’s claims. Shealso applies this description to in-depth interviews of 160 people on three different top-ics (crime, unemployment, and schools). For example, in the case of crime the issue was“What causes prisoners to return to crime after they’re released?” These explanations—or theories as she calls them—are coded for the number of causes cited and whetherthere is genuine evidence in support.

Kuhn’s real innovation is coding for and directly eliciting counterarguments. Counter-arguments suggest that people can envision the conditions that would falsify their expla-nations. This level of reasoning, especially if accompanied by genuine counterevidence,suggests a sophisticated knowledge of the topic well beyond that represented by reasonsand evidence for one’s own position. It indicates knowledge of alternative positions atodds with one’s own.

Kuhn’s interview protocol even moves to a third level of argumentative complexity—the ability to refute counterarguments to the original theory. Referring to the hypo-thetical person who offers a counterargument to the respondent’s original theory, theinterviewer asks “Would you be able to prove this person wrong?” and “What couldyou say to show that your own view is the correct one?” (Kuhn, 1991, p. 300).

The result of questions about arguments on an issue, counterarguments, and refuta-tions is a measure of the depth of knowledge held. Those who can identify multipleexplanations with genuine evidence for them, counterarguments to their own explana-tions, and a resolution in favor of their own explanation are at the highest levels ofknowledge about the issue under discussion.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 6: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

76 Joseph N. Cappella et al.

In our opinion, the importance of Kuhn’s work is not in its empirical findings butin its conceptual developments. Her theory offers a representation of knowledge that isat least a precondition for rational argument in face-to-face contexts. People with higherlevels of knowledge are contemplating the truth of their positions rather than simplyassuming the truth of their theories; that is, they are engaged in meta-cognition. Incontemplating the accuracy of one’s position, one is allowing the possibility that it couldbe wrong. As Kuhn (1991) notes,

paradoxically, this recognition is essential if one is to argue that theory iscorrect. . . . It is not primarily supporting evidence that shows a theory tobe correct. . . . One must also understand what evidence would show it to beuntrue and then seek out and confirm the absence of such evidence. Theability to conceive of counterarguments is thus fully as critical as the abilityto conceive of alternative theories. (p. 266)

Kuhn’s analysis of the nature of naive argument provides the conceptual outlines for anoperational indicator of opinion quality that (a) will be capable of ranking opinionsoffered by respondents in terms of their quality, (b) will be able to identify persons withthe tools to engage in deliberative discussion, and (c) can provide an indicator of theconsequences of deliberative discussion.

Kuhn’s conceptualization of argumentation dovetails with Park’s (2000) analysisof opinion in terms of components he names individuality and civility. Individualityrelates to the sophistication, consistency, and certainty of one’s personal views and,behaviorally, to one’s ability to argue those preferences assertively. Civility relates notto one’s own views but instead to one’s understanding of others. Both components, heargues, are necessary for effective democratic deliberation.

He interprets Habermas’s (1962/1989) theory of communicative action as empha-sizing mutual understanding and civility. A basic democratic trait, Park submits, is aware-ness of what other people think, coupled with some understanding of why others thinkthe way they do. Whereas speaking develops strength and individuality in opinion, it ishearing others speak that develops civility. Deliberation requires both: It is a “dual pro-cess of speaking and listening” (Park, 2000, p. 5). Put another way, what makes opiniondeliberative is not merely that it has been built upon careful contemplation, evidence,and supportive arguments, but also that it has grasped and taken into consideration theopposing view of others. Park identifies a range of cognitive, attitudinal, and behavioralaspects of civility. The cognitive elements of civility include breadth of understanding,perspective taking, and understanding others’ views. Attitudinal components includeempathy, tolerance, trust in others, and reciprocity. Behavioral components include speaking“with” (as opposed to speaking “for” oneself or “against” others), compromise, andconsensus building.

We focus here on what we take to be one of the core cognitive componentsof civility—namely, the consideredness of one’s opinion, both in the sense of havinganchored a viewpoint in argument and in the sense of having considered other, opposingviews. We propose “argument repertoires” (see Lustick & Miodownik, 2000), definedas the range of arguments people hold both in support of and against their favoredposition on a particular political issue or toward some political object. Our conceptualizationand measurement approach follows in some respects the consideredness measureused by Kim and colleagues (Kim, Wyatt, & Katz, 1999; Wyatt, Katz, & Kim, 2000).However, while their measure basically assesses the ability of respondents to argue their

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 7: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 77

views, our assessment takes direct empirical account of their understanding of others’points of view.

Previous research has employed versions of argument repertoire in the political arena.Woodard (1995) developed a measure of “argumentative depth” from Kuhn’s interviewprotocol using three levels: argument, counterargument, and refutation. He found themeasure to be reliable and to discriminate between those exposed and not exposed tonews reports about Clinton’s health care reform proposal. Rhee and Cappella (1997)found that political sophisticates exhibited higher quality arguments using a measureanalogous to Woodard’s. In both applications, relevant refutations—the third level—were very infrequent.

Defining Argument Repertoire

For any given stated opinion on an issue, argument repertoire includes the number ofrelevant reasons for the stated opinion and the number of relevant reasons for the oppo-site opinion. For example, if people indicate that they are favorable toward the Republi-can party, the number of relevant reasons in support of their opinion gives the first partof their repertoire score. They are then asked the reasons why someone might be un-favorable toward the Republican party. The number of such relevant reasons providesthe second part of their argument repertoire score.

The reasons provided must be relevant. By relevant, we mean reasons that areacknowledged in public discourse as plausible reasons (e.g., for their support of theRepublican party). Plausible reasons might include the Republicans’ policies regardingsmaller federal government and lowered taxes. Irrelevant reasons could include a person’sstatement that he or she liked Republicans or knew some Republicans. If irrelevantreasons were included, then argument repertoire would be little more than a word count,and no differentiation between genuine and bogus reasons and between reasons andmere preferences would be made.

Unlike Kuhn (1991), we do not try to include genuine evidence in support of rea-sons as a part of the evaluation of argument repertoire, nor do we attempt to ascertainthe coherence among multiple reasons (e.g., Tetlock, 1989). Both dimensions are plau-sible components of argument repertoire. However, we want to employ our measure in asurvey context. Genuine evidence is relatively infrequent, and Kuhn required a series ofquestions and probes in face-to-face interviews to elicit the little evidence that peopleoffered. Coherence (or integrative complexity) requires at least two arguments plus elaboration.In the current application, we have ignored integrative complexity.

Finally, our analytic system does not attempt to distinguish accurate from inaccu-rate reasons. For example, suppose a person favored the Republican party and did so“because the party’s policies favored the poor over the rich.” We do not judge thisclaim as inaccurate primarily because it is extremely difficult to know what reasons andevidence support the claim. A person could believe in trickle-down economics and sosupport policies that enrich the wealthy with the expectation of improving the plight ofthe poor. We leave factually accurate and inaccurate responses to the realm of standardclosed-ended knowledge questions.

Hypotheses

If argument repertoire is to be an effective measure of opinion quality, then it must beboth reliable and valid. Reliability is evaluated in terms of intercoder agreement (Krippendorff,

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 8: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

78 Joseph N. Cappella et al.

1980). The validity of argument repertoire should be established in terms of convergent,discriminate, and predictive validity. Tests of predictive validity depend on the theoreti-cal accounts within which argument repertoire participates.

Two tests of predictive validity are offered. The first hypothesizes that those high inargument repertoire will also be more likely to participate in subsequent group delibera-tions about the presidential campaign. The second hypothesizes that those participatingin group deliberations about the presidential campaign will show increases in subse-quent assessments of argument repertoire. These two hypotheses together describe aspiral of effects from argument repertoire to deliberation and from deliberation to argu-ment repertoire representative of the underlying conception that has given rise to theargument repertoire concept.

Method

Sample

Data are taken from parts of a multiwave survey of 1,684 adult Americans conductedduring the 2000 presidential election campaign. The study was a year-long panel consistingof surveys and discussions on-line. Respondents came from a random sample of Ameri-can citizens age 18 and older drawn from a nationally representative panel of surveyrespondents maintained by Knowledge Networks, Inc., of Menlo Park, California. TheKnowledge Networks panel includes a large number of households (in the tens of thou-sands) that have been selected through random digit dialing (RDD) and agreed to acceptfree WebTV equipment and service in exchange for completing periodic surveys on-line.1

The Electronic Dialogue 2000 project is distinguished from other Internet-basedstudies in a number of ways. While most studies examine asynchronous message boardsor less formal and happenstance “chat” experiences on the Web, this project createdsynchronous, real-time, moderated group discussions that were designed specifically toproduce useful citizen deliberation. Also, the project did not rely upon a conveniencesample of Internet users, as is common in Web-based studies; instead, it began with abroadly representative sample of Americans and attempted to recruit from that sample aset of 60 discussion groups that would be, in their entirety, as representative as possibleof U.S. citizens.

The project involved a multiwave, multigroup panel design, lasting roughly one year.All data gathering was conducted over the World Wide Web. The core of the projectconsisted of 60 groups of citizens who engaged in a series of monthly, real-time electronicdiscussions about issues facing the country and the unfolding presidential campaign.A set of baseline surveys in February and March 2000 assessed participants’ opinions,communication behaviors, knowledge of public affairs and of the presidential candi-dates, and a variety of demographic, personality, and background variables. Subsequentmonthly group deliberations began in April and generally included pre- and postdiscussionsurveys. The full text of all group discussions, which lasted an hour apiece, was recorded.A series of end-of-project surveys (and a final discussion) were then conducted in Januaryand February 2001. Further detail on the project, including a chronology of surveys,discussions, topics, and participation rates, is available in Price and Cappella (2001).

Measures

Argument repertoire. Argument repertoire was assessed for five separate topics duringthe year-long study. Here we report on three tests—one focused on political parties, one

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 9: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 79

on the candidates’ tax plans, and the third on the Supreme Court’s decision in Decem-ber 2000 in favor of George W. Bush’s petition to overturn the judgment of the FloridaSupreme Court. Analyses of the other topics (electoral college and choice for president)are in process.

At the baseline surveys in March 2000, all respondents were asked whether theywere favorable or unfavorable toward each of the two parties, on a 4-point ordinal scale.Following these questions, we asked them in four open-ended questions to write reasonswhy they were favorable or unfavorable toward the party and why other people wouldbe unfavorable or favorable toward it, respectively. (See the Appendix for exact ques-tion wording.)

Coding proceeded as follows.2 If the respondent left the question blank, he or shereceived a score of -99. If the respondent skipped the question, his or her answer wascoded as –2. If his or her answer was otherwise missing, the score was –1. When theanswer was irrelevant, did not make sense, merely restated the opinion, indicated thatthe person does not know why he or she holds that opinion, alluded in a vague way tothe parties’ positions, or was a statement about party membership only, it was coded aszero. For example, a statement such as “I like the Democratic party” or “The Democratssmell” was coded as zero. Except for those who registered no opinion on political party,all other skipped or blank responses were treated as zero.

For each substantive answer, one point was given for every reason the respondentwrote. For example,

Question: What are the reasons you have for feeling very unfavorable towardthe Republican party?

Answer: Views on abortion (1), too close ties to business interests (2), fightagainst raising minimum wage (3) and other ways to help the poor andworking class Americans (4), insistence on tax cuts (5), fight against makinghealth care benefits more available and affordable (6).

This answer received a score of 6.

Reasons for own opinion. In the example above, a person who is unfavorable toward theRepublican party is stating six different reasons why he or she is unfavorable, and there-fore that statement was coded as reasons for “own opinion.” Statements of both reasonswhy one favors the party with which one affiliates and reasons why one does not favorthe opposite party comprised this category. We constructed a combined index of thecoded responses, designating the sum of respondents’ reasons for holding their ownopinion ( a = .77, one dimension accounting for 81% of the variance). The number ofreasons for own opinion index ranged from 0 to 20, with about 16% giving no reasonsand about 46% giving 2 to 5 reasons (Mdn = 3, M = 3.94, SD = 3.25, N = 1,686).

Reasons why others might disagree. Following the questions about reasons for one’sown opinion, in an additional open-ended question, we asked the respondents to reasonwhy other people hold the opposite opinion. Someone who indicated that he or shewas favorable toward the Republican party was asked to name reasons why others mightbe unfavorable toward the Republican party. Similarly, a person stating she or he areunfavorable toward the Democratic party was asked to state reasons why others mightbe favorable toward the Democratic party. We constructed a combined index of the

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 10: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

80 Joseph N. Cappella et al.

coded responses, designating the sum of respondents’ reasons for the opinion of otherswho might disagree with their own position ( a = .80, one dimension accounting for83% of the variance). The index for the number of reasons for others’ opposite opinionranged from 0 to 16, with about 28% giving no reasons and slightly more than 40%giving 1 to 3 reasons (Mdn = 2, M = 2.66, SD = 2.64, N = 1,685).

Intercoder reliability was assessed on a subsample of 50 open-ended responses. Cohen’skappa values ranged between .70 and 1.00 (mostly in the .80 range).

Similar procedures were followed for Bush and Gore tax proposals assessed duringOctober 2000. Exact wording for these questions is included in the Appendix. Ownreasons summed across the two proposals ranged from 0 to 13 (Mdn = 2.0, M = 2.6,SD = 2.51, N = 566). Opponents’ reasons ranged from 0 to 8 (Mdn = 1, M = 1.6, SD =1.82, N = 566). Coding did not go forward until intercoder reliability produced kappasabove .75. Wording for the Supreme Court decision is included in the Appendix. Ownreasons summed ranged from 0 to 5 (Mdn = 1.0, M = 1.52, SD = 0.96, N = 967).Opponents’ reasons ranged from 0 to 7 (Mdn = 1, M = 1.13, SD = 0.91, N = 965 ).Intercoder reliability was above .70.

Political Knowledge

Various dimensions of political knowledge were combined to form a single scale mea-sure. Items included 10 general political and civics knowledge questions (e.g., Who hasthe final responsibility to decide if a law is constitutional or not?), 7 questions about thepersonal backgrounds of the presidential candidates (e.g., Which one of the democraticcandidates was a professional basketball player? Which one of the GOP candidates is aformer POW?), and an additional 7 questions about issue positions of candidates in theDemocratic and Republican presidential primaries (e.g., Which one of the Democraticcandidates supports universal health care? Which of the Republican candidates supportsvouchers?). All 24 items were scored 1 for correct answers and 0 for incorrect. Theitems were averaged to create a scale (Cronbach alpha = .82; M = .62, SD = .19).

Political Interest

We used two different items to construct a political interest scale. The questions, mea-sured on a 4-point ordinal scale, inquired about habitual following of public affairs andcaring which party wins in the 2000 elections. The majority of respondents (79%) re-ported that they followed public affairs either “most” or “some” of the time. About 50%of the respondents replied they cared “a great deal” which party wins the elections. Bothitems loaded on a single factor that explained 73% of the variance. A scale averagingthe two responses ( a = .62) was computed for further analyses (M = 3.20, SD = 0.71).

Mass Media Use

Exposure. Exposure to mass-mediated current events content was measured on five dif-ferent items inquiring about the respondents’ self-reported media use in days during thepast week (0 to 7). Newspaper reading and political talk radio exposure were measuredas single items. Three items—exposure to television national network news, cable news,and local news—were scaled together. A factor analysis of the three items yielded asingle factor explaining 59% of the variance in responses. A scale averaging the scores( a = .66) was computed (M = 3.46, SD = 1.93).

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 11: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 81

Attention. Two questions asked respondents to self-report their attention to articles aboutthe presidential campaign in newspapers and to reports about the campaign on televisionduring the past week. For each medium, responses were measured on a 5-point scaleranging from “a great deal” to “none.” Of the people who completed both waves of thebaseline survey, about 18% (n = 294) did not report any exposure to newspapers; 7.2%(n = 122) did not report exposure to television news. Newspaper attention ranged from0 to 5 (M = 2.49, SD = 1.52), as did television news attention (M = 3.04, SD = 1.34).

Political Conversation

Respondents were asked to name by initials up to four people with whom they regularlydiscussed politics, whether family members, friends, coworkers, acquaintances, or otherrelationships. In addition, respondents were asked to report several features of thesediscussions—among them, the frequency of talking about politics with these four per-sons and the extent of disagreement the main respondent perceived. Slightly less than11% did not name any discussion partners, and about 56% (n = 940) named four dis-cussion partners.

Frequency of discussion. The respondents reported how many days in the past week (0to 7) they discussed political issues with each of the four discussion partners. The re-sponses of those who did not name any discussants were coded as zero days theseindividuals discussed politics as well. An additive scale of the respondents’ answers tothe four questions was computed ( a = .74, M = 5.94, SD = 4.93).

Disagreement. For each of the four discussion partners, the respondents reported theextent to which they perceive their discussants to disagree with their own point of view.Disagreement was measured on a 5-point ordinal scale ranging from “never” to “almostall the time.” Survey respondents who did not have any discussion partners at all re-ceived a zero. An additive scale of the four disagreement scores ( a = .72) was com-puted for further analyses (M = 7.34, SD = 4.16).

Demographics

Respondents reported a median age of 43 years (M = 44.5, SD = 15.8); 46% werewomen, 54% men; 83% identified themselves as Caucasian, 7% as Black or AfricanAmerican, 2% as Asian, and 1.6% as Native American. Average education level was13.5 years (SD = 1.81), while 64% reported either full- or part-time employment.

Results

Descriptive (Baseline)

The questions assessing argument repertoire (AR) clearly are meaningful to respondents.On familiar topics such as reasons for favoring a political party, 5 of 6 people are ableto cite relevant reasons, while 3 of 4 can cite relevant reasons for why others are op-posed. On less familiar topics such as candidates’ tax plans, 7 in 10 can cite at least onerelevant reason for their own opinion, and 6 in 10 can cite a reason to have the oppositeopinion. On less familiar but extremely important topics such as the Supreme Court’selection decision, 9 of 10 offered at least one reason for their opinion and 3 of 4 areason for opponents’ positions.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 12: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

82 Joseph N. Cappella et al.

Procedures to assess reasons are reasonably well defined. Coders develop accept-able levels of reliability with relative ease (kappas above .75) and maintain them throughouttheir coding.

The two aspects of argument repertoire are not independent with own and others’reasons, correlating at .70, .74, and .46 for party, taxes, and Supreme Court ARs, re-spectively. These correlations suggest that those able to cite their own reasons are alsoable to cite opponents’ reasons. Yet, the two are not identical, with at least 50% of thevariance in others’ reasons not explained by own reasons.

Table 1 summarizes the associations between demographic characteristics and argu-ment repertoire. Across the three topics, those with the higher argument repertoire scoreshave higher educational achievement and more extreme scores on an ideology-partyindex, and they are more likely to be older, White, retired, and without children under18. The relationships to gender, ideology, and party identification are inconsistent. Thestrongest and most consistent findings are that those with anchored opinions are bettereducated and more committed to party and ideology.

Convergent Validity

One aspect of the validity of a measure is its correlation with variables that are sup-posed to tap into similar constructs. Table 2 presents correlations between argumentrepertoire for three issues and several constructs measuring aspects of the same latentdomain. The variables listed as rows are measured at the baseline survey. Argumentrepertoire for party is also measured at baseline, while argument repertoires for taxesand court are measured later. Thus, the correlations between subsequent argument reper-toire measures and earlier, transient variables such as attention to news, network size,and disagreement with discussion partners are not included in the table.3

Table 1Correlations between argument repertoire scales and demographic variables

Party Tax Court

Own Other Own Other Own Other(N = 1,628) (N = 1,627) (N = 563) (N = 563) (N = 921) (N = 920)

Others’ reasons .70*** .74*** .46***More educated .24*** .29*** .16*** .20*** .15*** .21***Conservative ideology .10** .04 .03 .03 .07† .07More Republican .07** .02 .06† .05 .02 .02Extremisma .24*** .14*** .18*** .17*** .08* .10***Male .08** .05* –.02 .03 –.04 –.01White .09** .10*** .07† .05 .05 .10***Older .16*** .14*** .18*** .12*** .08* .09**Retired .07** .09** .11** .07† .02 .03No children under 18 .06* .06* .13** .11** –.02 -.05

Note. Own refers to number of reasons for one’s own position. Other refers to reasons for those withopposing position.

aAbsolute value of a ten-point scale (–5 to 5) of party affiliation and ideology.†p < .10; *p < .05; **p < .01; ***p < .001.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 13: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 83

Those higher in argument repertoire also tend to be more interested in politics,more knowledgeable about politics, more attentive to political news, and more likely toread newspapers for news. Moreover, they have larger political discussion networks andperceive these groups to have more opposed positions.

Also shown in Table 2 is a correlation with “flexibility,” a variable that calculates aparticipant’s availability to join proposed discussion groups at scheduled times. The higherthe flexibility, the more free time available to the respondent. One possibility is thatthose with a great deal of free time might also be those who are able to write longerresponses to our open questions about their own and others’ reasons. The results inTable 2 are mixed on this question, showing 4 of 6 significant, positive correlations. Weinvestigated this relationship more carefully in a series of regressions not reported here.With argument repertoire as the outcome variable and flexibility, age, retirement status,children in the home, political interest, education, and political knowledge in the equa-tion, flexibility drops out as a significant predictor, while political knowledge, interest,and education remain. This suggests that simple availability of time is less a reason forthe higher argument repertoire scores than is motivation to answer.

The correlations in Table 2 are in line with our expectations about how argumentrepertoire should function. Those with the capacity to write out reasons for their opinionsand to identify relevant reasons for opposed opinions also express interest in politics,are more accurate in their factual political knowledge, and use the print and broadcastmedia as sources of their political news. Even their personal communication is more

Table 2Convergent validity argument repertoire scales: Bivariate correlations

Party Tax Court

Own Other Own Other Own Other(N = 1,629) (N = 1,628) (N = 563) (N = 563) (N = 921) (N = 920)

Political knowledge .48*** .46*** .29*** .35*** .21*** .27***Political interesta .35*** .30*** .27*** .24*** .12*** .13***Flexibility .13*** .14*** .13** .09* .05 .03Newspaper exposure .12*** .12*** .15*** .13** .08* .12***Television news .05* .02 .15*** .02 .00 .02

exposurePolitical talk radio .14*** .14*** .03 .04 .05 .02

exposureAttention to campaign

Newspapers .23*** .22***Television news .24*** .20***

Network size .30*** .30***Perceived disagreement .22*** .25***

of discussion partners

Note. Own refers to number of reasons for one’s own position. Other refers to reasons for those withopposing position.

aMean interest scale, averaging interest in public affairs and how much the individual cares whichparty wins the 2000 election (1–4 scale).

*p < .05; **p < .01; ***p < .001.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 14: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

84 Joseph N. Cappella et al.

political and more diverse (see Price, Cappella, & Nir, 2002). Coupled with the datafrom Table 1 indicating higher argument repertoire for those with more education andmore commitment to their ideology and party, we have good evidence of convergentvalidity.

Consistency

A person’s ability to provide reasons for his or her own and opponents’ positions mightvery well change as issues change. With different issues, familiarity, exposure to news,and interest may also differ. However, some stability in argument repertoire scores shouldcharacterize this form of political knowledge. In Table 3, we offer the correlations be-tween points in time for three different measures of argument repertoire. Argument rep-ertoire at baseline (political parties) is compared with AR (for tax proposals and theSupreme Court decision). The AR for tax was taken 7 months later and that for thecourt decision 10 months later.

Despite these lengthy elapsed times, all correlations are significant and of moderatesize. The correlations between AR party and AR court are lower than those for partyand taxes. One possible reason is that the party and taxes measures are actually sumsacross two different questions (i.e., both political parties and both candidates’ tax plans).AR court is based on only one question regarding reasons for favoring or opposing theSupreme Court’s decision and one question for opponents’ reasons for the oppositeposition.

Nevertheless, despite the elapse of long periods of time, slightly different measure-ment procedures, and very different issues, those with higher AR scores tend to havehigher scores on other issues at later points in time.

Predictive Validity (Consequences of Argument Repertoire)

Those with elevated AR scores should be people with the capacity to participate inpolitical deliberation. Their argument repertoires give them not just factual knowledgebut knowledge that can be used in deliberative activity (i.e., knowledge of their ownopinions and of the opinions of their interlocutors). Thus, AR should predict willingness

Table 3Consistency of argument repertoire scores over time and topic: Correlations

Tax proposals Supreme Court decision(October 7–18, 2000) (December 4–15, 2000)

(N = 563) (N = 920)

Own Opponents’ Own Opponents’reasons reasons reasons reasons

Own reasons: party .42 .43 .29 .31(March 10–23, 2000)

Opponents’ reasons: party .39 .41 .28 .34(March 10–23, 2000)

Note. All correlations are significant at p < .001.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 15: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 85

to participate in deliberative group discussions about politics and, once there, delibera-tive behaviors contributing to group exchange.

In Table 4, we present results from multivariate logistic regressions predicting par-ticipation in any of the events. 4 Of the demographic variables, only age predicts partici-pation after controls, such that older respondents are more likely to attend.5 Time con-straints appear to influence participation in a number of ways. Having children at home,being employed full time, and being a full-time student were each negatively associatedwith participation. The effect of time constraints is further demonstrated by the measureof schedule flexibility. The more flexible respondents (those who said they were avail-able for participation in more time slots during our preparation of group assignments)were considerably more likely to attend the project’s electronic events.

Political knowledge, political discussion, community participation, and interpersonaltrust also affected participation significantly (and positively, as expected), controllingfor time constraints, ideological tendencies, and demographic factors. The higher peoplescored on each of these measures, the higher the estimated odds that they would partici-pate in at least one of the electronic discussions. Overall, the pattern of coefficientsstrongly supports the view that “social capital” goes hand in hand with political partici-pation (e.g., Putnam, 2000). Meanwhile, mass media use had no such effect. In fact,after application of these extensive controls, it is actually the case that people who

Table 4Logistic regressions predicting discussion attendance from argument repertoire

Model 1 Model 2

B eb B eb

Sex (male = 1) –.17 .84 –.16 .86Race (White = 1) –.02 .94 –.05 .95Years of education .03 1.03 .02 1.02Age .03*** 1.03 .03*** 1.03Party-ideology index –.01 .99 –.02 .99News media exposure –.15*** .86 –.14** .87Political knowledge 1.10* 3.00 .15 1.15Political discussion .04 1.04 .02 1.01Interpersonal trust .62** 1.86 .47† 1.61Community participation 1.23*** 3.41 1.05*** 2.85Schedule flexibility .18*** 1.20 .17*** 1.18Employed full time –.53*** .59 –.48*** .62Number of children –.14† .87 –.16† .86Full time student –1.34** .26 –1.44** .24Argumentation .11*** 1.14

R2 .21 .22

Cases correctly classified (%) 75 75

N 810 778

Note. The dependent variable is attendance in at least one online discussion.†p < .1; *p < .05; **p < .01; ***p < .001.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 16: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

86 Joseph N. Cappella et al.

report higher levels of exposure to newspapers, TV, and talk radio were somewhat lesslikely to turn out for their on-line discussions.

Model 2 in Table 4 adds to the equation argument repertoire. This measure capturesthe effects of political knowledge and political conversation, which no longer producesignificant coefficients, and weakens somewhat the impact of interpersonal trust. Evenafter extensive controls, AR for political parties measured at baseline accounts forattendance in subsequent on-line deliberations. Consistent with our expectations, thosewith elevated AR scores tend to be more willing to participate in subsequent politicalconversations.

Willingness to attend on-line group discussions is not the same as contributing tothose discussions. In Table 5, we present correlations between AR for political partiesassessed at baseline and behavioral participation in a discussion of education two monthslater. In all of our discussions, an initial period was devoted to welcoming participantsas they logged in. This short period included general chitchat. The remainder of theconversation was directed toward substantive issues determined a priori and led by amoderator. In this particular event, the preponderance of arguments went against vouch-ers as a solution to problems in the schools (Price & Cappella, 2001). Those higher inAR generated more words, more substantive words on the central topics of the conver-sation, and, most important, more arguments against vouchers.

AR shows predictive validity in two important senses. Those with greater ARs tendto be more willing to attend group deliberations about political topics and they tend toparticipate in the substantive exchanges while there. They put their argumentative skillto use.

Predicting Argument Repertoire

In October, our participants discussed the tax proposals advanced by candidates Bushand Gore. In the survey completed in the weeks following, respondents were asked theirviews of both candidates’ tax plan. If they stated an opinion, argument repertoire scoreswere obtained for each candidate’s plan.

As shown in the top panel of Table 6, the number of discussions attended by projectparticipants significantly predicts scores on this argument repertoire measure, controlling

Table 5Correlations between argument repertoire at baselineand subsequent communication during deliberation

Communication duringeducation discussion Own reasons: party Opponents’s reasons: party(May 5–21, 2000) (March 10–23, 2000) (March 10–23, 2000)

Total words .14* .22**Words: vouchers .16** .20***Words: character education .21*** .25***Pro arguments: vouchers .09 .07Con arguments: vouchers .18** .18**

Note. *p < .05; **p < .01; ***p < .001.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 17: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 87

for argument repertoire assessed on the baseline survey and for propensity to attend.Furthermore, it appears to be attendance at the October event in particular—when taxeswere discussed—that is primarily responsible for this effect. As shown in the lowerpanel of Table 6, an equation that includes dummies for participation at multiple events(e.g., the May event, when education was discussed, as well as the October event) pro-duced significant coefficients only for the most relevant event, when the candidates’ taxproposals were discussed.

Argument repertoire is affected by deliberative group interactions. Both the act ofdeliberating and the substance of conversations can be efficacious in elevating argumentrepertoire. The details of how this effect is produced remain to be uncovered. However,

Table 6Regressions predicting tax proposal argument

repertoire from participation in discussions

Own reasons Others’ reasons Total reasons

B b B b B b

Regressions using number of events attended as predictor

Baseline Argument .27*** .37 .24*** .36 .30*** .43repertoire

Number of discussions .11* .09 .09** .10 .18* .09attended (throughOctober)

Propensity to attend 2.08** .12 1.40** .11 2.92* .10Intercept –.27 –.18 –.41

R2 .21 .18 .24

N 527 525 525

Regressions using specific events attended as predictors

Baseline argument .27*** .37 .23*** .36 .30*** .43repertoire

Attended in May –.02 .00 .22 .06 .18 .02(education discussed)

Attended in October .56** .11 .14 .04 .66* .08(taxes discussed)

Propensity to attend 2.01** .12 1.48** .12 2.89* .10Intercept –.20 –.15 –.32

R2 .21 .19 .26

N 527 525 525

Notes. Entries are ordinary least squares regression coefficients. Dependent variables are countsof the number of reasons coded as relevant to opinions on the candidates’ tax proposals.

*p < .05; **p < .01; ***p < .001

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 18: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

88 Joseph N. Cappella et al.

the AR measure itself is sensitive enough to pick up differences resulting from topicaldeliberations.

Discussion

The research reported in this paper offers a new measure tapping one component ofopinion quality, a component representing the degree of anchoring of one’s own opinionand knowledge of the bases for opinions held by others. These two aspects are pre-sumed to tap into the ability to express one’s own viewpoints and to be open to alterna-tive viewpoints. In short, those with more extensive argument repertoires should also becapable of participating effectively in deliberative discussions, and opinion quality shouldbe a sensitive measure of the impact of substantive give-and-take about politics fromdeliberative groups. These claims have been tested in the research reported here.

Argument repertoire is a reliable measure of opinion quality. In three applications,coders were able to make reliable assessments of the relevant reasons people have fortheir opinions and reasons that others have for holding opposed opinions. Coding pro-cedures are relatively straightforward and do not require extensive training time. Re-spondents treat the questions assessing AR as meaningful, with substantial proportionsgiving at least one relevant reason for their own and for opposed opinions. AR hassome stability across time and across topics. While topical variation is expected, ARwould be a less effective measure if different topics produced totally different AR scores.

AR is also a valid measure of opinion quality. Those with higher scores on AR arepeople with expected political and communication profiles. Those with higher AR aremore educated, have greater political knowledge, more interest in politics, more expo-sure and attention to news, more interpersonal communication about politics, more com-mitment to their political parties, and are older. People we expect to have higher levelsof AR have higher levels of AR.

AR shows itself to be an effective indicator of participation in political discussion.Respondents with higher ARs were more likely to participate in our on-line discussiongroups and, once there, talked more on topic, including offering arguments about the issueunder consideration. The AR measure has proved to be sensitive to the effects of delib-eration, adding to participants’ prior levels of reasons for their own and others’ opinions.

Together, these two effects suggest a spiral between deliberative discussion andAR, with each being a causal force in the other’s growth at a later point in time.

The research we have completed so far has not yet addressed two key questions:the bases of argument repertoire and argument quality. Although we have seen thatdeliberation can increase AR, the mechanism for this change is not yet well understood.The availability of divergent opinion within the group’s discussion is a plausible candi-date (Price, Cappella, & Nir, 2002). One’s own reasons may grow as one searches forresponses to challenges from others. Knowledge of others’ opinions may increase asmore diverse reasons are heard from others. On the negative side, disagreement mayproduce polarization and clique formation within heterogeneous groups. Understandingthe processes through which ARs develop will help us understand the role of delibera-tion in developing informed citizens.

We have been careful not to label our measure argument quality. Quality impliesthat the reasons coded are somehow superior to other reasons offered or that evidence ismore serious than other evidence considered. The quality of arguments that people haveat their disposal at a minimum should take into account the relevance, number, integra-tion (or coherence), and accuracy of reasons offered. Our measure of AR includes only

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 19: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 89

the first two components of argument quality. Although we have presented consistentand solid correlations between AR and factual political knowledge, accuracy and coher-ence remain challenges for future research with AR.

AR holds considerable promise for future research. It can be used to distinguishdeeper from more shallow opinions, and in this sense it is a component of opinionquality. Moreover, the concept is built around the idea that information in the form ofreasons for one’s own and for others’ positions is central to opinions that are deliberative—opinions that can be the focus of communicative give-and-take. Without a sense ofone’s own reasons and without a sense of the audience’s, political communication asdeliberation with others is undermined.

Notes

1. The Knowledge Networks panel sample begins with a list-assisted RDD sampleprovided by Survey Sampling, Inc. (SSI). Samples are acquired approximately once amonth to ensure that they are drawn from up-to-date databases. Numbers in the SSIsample are then matched against a database of numbers known to be in the WebTVnetwork. These numbers are then contacted, and households are asked to participate asmembers of the Knowledge Networks panel. In exchange for completing surveys (ap-proximately 40 minutes of cumulative survey time per household per month), panelistsreceive WebTV equipment and access free of charge. The recruitment process results ina response rate of approximately 55% to 60%. It produces a sample of American house-holds that closely approximates the population at large, with a very slight underrepresen-tation of minorities and the elderly (Knowledge Networks, 2000). In February 2000, arandom sample of American citizens age 18 and older (N = 3,967) was drawn from theKnowledge Networks panel, with the intention of recruiting participants for the Elec-tronic Dialogue 2000 project. The initial, recruitment survey provided a brief descriptionof the project, emphasized the need for a representative sample, and included a formindicating a respondent’s consent to participate. Overall, 51% of those recruited agreedto participate. Those who consented were then sent two baseline surveys, the first fromFebruary 8 to March 10 and the second from March 10 to March 23. The surveysincluded extensive measures of media use, interest in the presidential campaign, generalpolitical knowledge and knowledge of the campaign, political discussion, and a widevariety of political attitudes and opinions. Response rates to each of the baselines wereapproximately 90%.

2. A detailed coding manual is available from the authors upon request.3. When these correlations are calculated, they exhibit the same pattern of results,

only slightly weaker in magnitude.4. Other analyses (not shown) examined predictors of the number of discussions

attended rather than the odds of attending any single event. Results are quite similar.5. Several of the political variables were not significantly associated with on-line

attendance and hence are not presented in Table 4. These include political efficacy,political interest, and political participation. These variables do not significantly predictparticipation in the electronic discussion after controlling for demographics and politicalmeasures in Table 4.

References

Arendt, H. (1958). The human condition. Chicago: University of Chicago Press.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 20: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

90 Joseph N. Cappella et al.

Bishop, G. F., Oldendick, R. W., Tuchfarber, A. J., & Bennett, S. E. (1980). Pseudo-opinions onpublic affairs. Public Opinion Quarterly, 44, 198–209.

Cappella, J. N., & Folger, J. P. (1980). An information processing explanation of the attitude-behavior inconsistency. In D. P. Cushman & R. McPhee (Eds.), Message-attitude-behaviorrelationship: Theory, methodology, and application (pp. 149–194). New York: Academic Press.

Converse, J. M., & Presser, S. (1986). Survey questions: Handcrafting the standardized question-naire. Beverly Hills, CA: Sage.

Converse, P. E. (1964). The nature of belief systems in mass publics. In D. E. Apter (Ed.), Ideol-ogy and discontent (pp. 206–261). New York: Free Press.

Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it mat-ters. New Haven, CT: Yale University Press.

Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introduction to theoryand research. Boston: Addison-Wesley.

Fishkin, J. S. (1991). Democracy and deliberation: New directions for democratic reform. NewHaven, CT: Yale University Press.

Fishkin, J. S. (1995). The voice of the people: Public opinion and democracy. New Haven, CT:Yale University Press.

Gutmann, A., & Thompson, D. (1996). Democracy and disagreement. Cambridge, MA: HarvardUniversity Press.

Habermas, J. (1984). The theory of communicative action, Vol. 1. (T. McCarthy, Trans.). Boston:Beacon. (Original work published 1981)

Habermas, J. (1989). The structural transformation of the public sphere: An inquiry into a cat-egory of the bourgeois society (T. Burger, Trans.). Cambridge, MA: MIT Press. (Originalwork published 1962)

Herbst, S. (1993). Numbered voices: How opinion polling has shaped American politics. Chicago:University of Chicago Press.

Kay, A. F., Henderson, H., Steeper, F., & Lake, C. (1994). Interviews with the public guide us . .. on the road to consensus. St. Augustine, FL: Americans Talk Issues Forum.

Kim, J., Wyatt, R. O., & Katz, E. (1999). News, talk, opinion, participation: The part played byconversation in deliberative democracy. Political Communication, 16, 361–385.

Knowledge Networks. (2000). Sampling report: Annenberg survey. Unpublished report.Krippendorff, K. (1980). Content analysis: An introduction to its methodology. Beverly Hills, CA:

Sage.Kuhn, D. (1991). The skills of argument. New York: Cambridge University Press.Lustick, I., & Miodownik, D. (2000). Deliberative democracy and public discourse: The agent-

based argument repertoire model. Complexity, 5(4), 13–30.Neijens, P. (1987). The choice questionnaire: Design and evaluation of an instrument for collect-

ing informed opinions of a population. Amsterdam: Free University Press.Neijens, P., de Ridder, J. A., & Saris, W. E. (1992). An instrument for collecting informed opin-

ions. Quality and Quantity, 26, 245–258.Neuman, W. R. (1986). The paradox of mass politics: Knowledge and opinion in the American

electorate. Cambridge, MA: Harvard University Press.Park, S. G. (2000, October). The significance of civility in deliberative democracy. Paper pre-

sented at the conference of the Public Opinion Research in the Digital Age (PORDA) project,Seoul, Korea.

Pratkanis, A. (1989). The cognitive representation of attitudes. In A. Pratkanis, S. Breckler, & A.Greenwald (Eds.), Attitude structure and function. Hillsdale, NJ: Erlbaum.

Price, V., & Cappella, J. N. (2001, May). Online deliberation and its influence: The ElectronicDialogue project in campaign 2000. Paper presented at the meeting of the American Asso-ciation for Public Opinion Research, Montreal, Quebec, Canada.

Price, V., Cappella, J. N., & Nir, L. (2002). Does disagreement contribute to more deliberativeopinion? Political Communication, 19, 97–114.

Price, V., & Neijens, P. (1997). Opinion quality in public opinion research. International Journalof Public Opinion Research, 9, 336–360.

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 21: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 91

Price, V., & Neijens, P. C. (1998). Deliberative polls: Toward improved measures of informedpublic opinion? International Journal of Public Opinion Research, 10, 145–176.

Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. NewYork: Simon & Schuster.

Rhee, J. W., & Cappella, J. N. (1997). The role of political sophistication in learning from news:Measuring schema development. Communication Research, 24, 197–233.

Rosen, J. (1991). Making journalism more public. Communication, 12, 267–284.Schuman, H., & Presser, S. (1981). Questions and answers in attitude surveys: Experiments on

question form, wording, and context. New York: Academic Press.Tetlock, P. (1989). Structure and function in political belief systems. In A. Pratkanis, S. Breckler,

& A. Greenwald (Eds.), Attitude structure and function. Hillsdale, NJ: Erlbaum.Woodard, E. (1995, May). Argumentative skill: A measure of schema development. Paper pre-

sented at the meeting of the International Communication Association, Albuquerque, NM.Wyatt, R. O., Katz, E., & Kim, J. (2000). Bridging the spheres: Political and personal conversa-

tion in public and private spaces. Journal of Communication, 50, 71–92.Zaller, J. R. (1992). The nature and origins of mass opinion. Cambridge, England: Cambridge

University Press.

Appendix: Question Wording for Argument Repertoire

Argument Repertoire: Baseline (March 10–23, 2000)

Reasons for Own Opinion and Reasons Why Others Might DisagreeR1. How favorable in general are you toward the Democratic party?

1. Very favorable2. Somewhat favorable3. Somewhat unfavorable4. Very unfavorableAsk R1a-b if favorable [R1(1-2)]

R1a. What are the reasons you have for feeling (very/somewhat) favorable toward theDemocratic party? (Please list all the reasons that come to mind) [textbox]

R1b. What reasons do you think other people might have for feeling unfavorable to-ward the Democratic party? (Please list all the reasons that come to mind) [textbox]Ask R1c-d if unfavorable [R1(3-4)]

R1c. What are the reasons you have for feeling (very/somewhat) unfavorable towardthe Democratic party? (Please list all the reasons that come to mind) [textbox]

R1d. What reasons do you think other people might have for feeling favorable towardthe Democratic party? (Please list all the reasons that come to mind) [textbox]

R2. How favorable in general are you toward the Republican party?Very favorable1. Somewhat favorable2. Somewhat unfavorable3. Very unfavorableAsk R2a-b if favorable [R2(1-2)]

R2a. What are the reasons you have for feeling (very/somewhat) favorable toward theRepublican party? (Please list all the reasons that come to mind) [textbox]

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 22: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

92 Joseph N. Cappella et al.

R2b. What reasons do you think other people might have for feeling unfavorable to-ward the Republican party? (Please list all the reasons that come to mind) [textbox]

Ask R2c-d if unfavorable [R2(3-4)]

R2c. What are the reasons you have for feeling (very/somewhat) unfavorable towardthe Republican party? (Please list all the reasons that come to mind) [textbox]

R2d. What reasons do you think other people might have for feeling favorable towardthe Republican party? (Please list all the reasons that come to mind) [textbox]

Argument Repertoire Regarding Tax Proposals(October 7–18, 2000)

In their campaigns for President, the candidates have made different proposals for cut-ting federal taxes. We’d like to know some of your reactions.

[Randomize Order of C1 and C2]

C1. How favorable in general are you toward the tax proposals that have been madeby Al Gore?[radio]1. Very favorable2. Somewhat favorable3. Somewhat unfavorable4. Very unfavorableAsk C1a-b if favorable [C1(1-2)]

C1a. What are the reasons you have for feeling (very/somewhat) favorable towardGore’s tax proposals? (Please list all the reasons that come to mind) [textbox]

C1b. What reasons do you think other people might have for feeling unfavorabletoward Gore’s tax proposals? (Please list all the reasons that come to mind)[textbox]

Ask C1c-d if unfavorable [C1(3-4)]

C1c. What are the reasons you have for feeling (very/somewhat) unfavorable towardGore’s tax proposals? (Please list all the reasons that come to mind) [textbox]

C1d. What reasons do you think other people might have for feeling favorable towardGore’s tax proposals? (Please list all the reasons that come to mind) [textbox]

C2. How favorable in general are you toward the tax proposals that have been madeby George W. Bush?[radio]1. Very favorable2. Somewhat favorable3. Somewhat unfavorable4. Very unfavorableAsk C2a-b if favorable [C2(1-2)]

C2a. What are the reasons you have for feeling (very/somewhat) favorable towardGeorge W. Bush’s tax proposals? (Please list all the reasons that come to mind)[textbox]

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4

Page 23: Argument Repertoire as a Reliable and Valid Measure of Opinion Quality: Electronic Dialogue During Campaign 2000

Argument Repertoire 93

C2b. What reasons do you think other people might have for feeling unfavorable to-ward George W. Bush’s tax proposals? (Please list all the reasons that come tomind) [textbox]

Ask C2c-d if unfavorable [C2(3-4)]

C2c. What are the reasons you have for feeling (very/somewhat) unfavorable towardGeorge W. Bush’s tax proposals? (Please list all the reasons that come to mind)[textbox]

C2d. What reasons do you think other people might have for feeling favorable towardGeorge W. Bush’s tax proposals? (Please list all the reasons that come to mind)[textbox]

Argument Repertoire Questions for Supreme CourtDecision (January 4–18, 2001)

C1. As you may be aware, in December the U.S. Supreme Court ruled, by a 5–4majority, in favor of George W. Bush and overturned the Florida Supreme Court’scall for a statewide, manual recount of disputed election ballots in Florida. Didyou support or oppose the U.S Supreme Court decision that ended the recount?[radio]1. Support strongly2. Support somewhat3. Oppose somewhat4. Oppose strongly

If C1 = 1 or 2

C1a. What are your reasons for supporting the Court’s decision? (Please list all thereasons that come to mind) [textbox]

If C1 = 1 or 2

C1b. What reasons do you think other people might have for opposing the Court’sdecision? (Please list all the reasons that come to mind) [textbox]

If C1 = 3 or 4

C1c. What are your reasons for opposing the Court’s decision? (Please list all the rea-sons that come to mind) [textbox]

If C1 = 3 or 4

C1d. What reasons do you think other people might have for supporting the Court’sdecision? (Please list all the reasons that come to mind) [textbox]

Dow

nloa

ded

by [

Was

hing

ton

Stat

e U

nive

rsity

Lib

rari

es ]

at 1

6:35

11

Oct

ober

201

4