Deference, Dissent, and Dispute Resolution: An ... · Deference, Dissent, and Dispute Resolution November 2009 The article is structured as follows. We begin by discussing media interventions

American Political Science Review Vol. 103, No. 4 November 2009

doi:10.1017/S0003055409990128

Deference, Dissent, and Dispute Resolution:An Experimental Intervention Using Mass Mediato Change Norms and Behavior in RwandaELIZABETH LEVY PALUCK Princeton UniversityDONALD P. GREEN Yale University

Deference and dissent strike a delicate balance in any polity. Insufficient deference to authoritymay incapacitate government, whereas too much may allow leaders to orchestrate mass violence.Although cross-national and cross-temporal variation in deference to authority and willingness

to express dissent has long been studied in political science, rarely have scholars studied programsdesigned to change these aspects of political culture. This study, situated in post-genocide Rwanda, reportsa qualitative and quantitative assessment of one such attempt, a radio program aimed at discouragingblind obedience and reliance on direction from authorities and promoting independent thought andcollective action in problem solving. Over the course of one year, this radio program or a comparableprogram dealing with HIV was randomly presented to pairs of communities, including communities ofgenocide survivors, Twa people, and imprisoned genocidaires. Changes in individual attitudes, perceivedcommunity norms, and deliberative behaviors were assessed using closed-ended interviews, focus groupdiscussions, role-play exercises, and unobtrusive measures of collective decision making. Although theradio program had little effect on many kinds of beliefs and attitudes, it had a substantial impact onlisteners’ willingness to express dissent and the ways they resolved communal problems.

Societies in which genocides occur are often saidto have distinctive political cultures in which def-erence to authority and conformity run high.

Whether obedience is traced theoretically to child-rearing practices (Adorno et al. 1950), cultures withinlarge bureaucratic organizations (Arendt 1963), highlycentralized authority (Chalk and Jonassohn 1990),or individuals’ motivation to reduce internal dis-sonance (Hinton 2005), genocidal perpetrators andfellow-travelers in Germany (Craig 1980; Lewin 1948),Cambodia (Chandler 1999), and Turkey (Ramsauer1957) are said to be unusually compliant and unusuallyreluctant to express dissent.

The cultural origins of genocide have special reso-nance for the interpretation of the Rwandan genocidein 1994 and for efforts to develop interventions to pre-

Elizabeth Levy Paluck is Assistant Professor of Psychology and Pub-lic Affairs, Department of Psychology, Princeton University, GreenHall, Princeton, NJ 08540 ([email protected]).

Donald P. Green is A. Whitney Griswold Professor of PoliticalScience and Director of the Institution for Social and Policy Studies,Yale University, 77 Prospect Street, Room 106, P.O. Box 208301, NewHaven, CT 06520-8301 ([email protected]).

The authors acknowledge the indispensable research assistanceof Albert Nzamukwereka and the team of 21 Rwandan researchers,in particular, Willy Kamanzi and Virgile Uzabumugabo. We thankLaBenevolencija, in particular, Johan Deflander, Suzanne Fisher,Laurie Pearlman, Ervin Staub, and George Weiss. John Bullock,Desha Girod, Devra C. Moehler, Christopher Muller, JeremyWeinstein, Pedro Vicente, the CISAC seminar at Stanford, and theHarvard Civil Conflict Seminar provided valuable comments. Wethank Pamela Lamonaca, Peter Aronow, and Celia Paris for theirassistance in preparing the manuscript. This research was supportedby the John Enders Fund, a Yale Center for the Study of Globaliza-tion Research Grant, the John Perry Miller Fund Prize, a NationalScience Foundation Graduate Fellowship to Elizabeth Levy Paluck,LaBenevolencija, and the Institution for Social and Policy Studies atYale University.

vent future ethnic violence. Journalistic accounts stressRwanda’s “entrenched culture of obedience” whenaccounting for the swiftness with which Hutus killedtheir Tutsi victims (Lacey 2004). A scholarly history ofthe genocide likewise describes the Rwandan citizen’smind-set: “[w]hen the highest authorities in that statetold you to do something, you did it, even if it includedkilling” (Prunier 1995, 245). Political and human rightsobservers concur (African Rights 1995, 249; Dallaireand Beardsley 2003; Prendergast and Smock 1999, 13).Rural Rwandan interviewees have commented to var-ious academic researchers that Rwandans are meekfollowers, “like cows” (Lyons and Straus 2006; Paluck2006; Zorbas 2007). The current Tutsi-led government,moreover, decries the “[p]assivity, lack of initiative anddependency syndrome on the part of the majority ofthe population” (Republic of Rwanda 2001, 4).

The adequacy of such cultural explanations of geno-cide is debatable.1 Nevertheless, many post-conflictpolicies and programs target culture, which is be-lieved to exacerbate intergroup enmity and facili-tate structural problems that perpetuate antagonism.Governments and international and nongovernmental

1 The cultural traits that are said to facilitate genocide often fail to doso unless accompanied by an array of contextual factors. Obediencedepends on many situational factors such as the presence of peers,inconsistencies in authoritative instructions (Milgram 1974; Packer2008), and the support of like-minded dissenters (Asch 1956; Bondand Smith 1996). Add to this the many proximal causes of genoci-dal violence, such as small-group dynamics favoring peer pressure,solidarity, and conformity (Browning 1993; Fuji 2009; Straus 2006),the beliefs and attitudes that subordinates harbor about their vic-tims (Goldhagen 1996; Klemperer 1998), and economic opportunism(Andre and Platteau 1998). The contextual emphasis of this researchreveals the inadequacy of cultural explanations involving static ac-counts of national character (Almond and Verba 1963; Eckstein 1992;Huntington 2000; Inglehart 1997; Parsons 1951; Pye and Verba 1965).

622

American Political Science Review Vol. 103, No. 4

organizations have embarked on ambitious culturalchange campaigns to address violence, corruption, andother post-conflict problems (e.g., Transparency Inter-national 2008; United Nations 1999). The governmentof Rwanda, for example, officially promotes develop-ment policies incorporating “open debate with collec-tive responsibility for development . . . to replace theculture of passive obedience” (Republic of Rwanda2002, 4).

These efforts raise fundamental questions about theextent to which, and the conditions under which, polit-ical culture changes. To answer this question, one mustfirst define political culture and explain how it guidespolitical conduct. Although there is little consensus onthis matter, Sewell (2005) argues that contemporarywork on culture falls into two general categories. Thefirst characterizes culture as a broad system of sharedmeanings and symbols (e.g., Laitin 1986; Schatzberg2001).2 The second characterizes culture largely asbehavioral practice, stressing forms of conduct thatconstitute an individual’s behavioral repertoires or“toolkit” (Swidler 1986; see Wedeen 2002).3 Drawingon Sewell (2005), we treat political culture as neitherone nor the other; for us, culture is a system of meaninglinked to a set of available behavioral practices. Forour purposes, we focus on the practices believed tobe most closely related to genocide—deference to au-thority, willingness to speak out in defiance of a groupnorm, and acceptance of collective responsibility forthe mediation of disputes.4

Shifting the locus of culture away from national char-acter to everyday behavior allows for the possibilitythat interventions can change culture by targeting po-litical conduct. Sewell notes that the relationship be-tween system and practice “[possesses] a real but thincoherence that is continually put at risk in practiceand therefore subject to transformation” (2005, 169).However, most observational research on political cul-tural change has examined change through shifts inlarge-scale forces, to the neglect of more modest inter-ventions aimed at affecting political conduct. For ex-ample, Rogowski (1974) argues that German politicalattitudes shifted dramatically after World War II, withrelatively muted differences between age cohorts so-cialized under the Third Reich and those who came ofpolitical age afterward (see also Laitin 1986 on Britishcolonial rule). Likewise, Rwandans’ deference to au-thority has been traced to institutional arrangementsimposed by Belgian colonial rulers, who refashioned

2 Other research that portrays political culture as a shared, coherentsystem defines it differently, for example, as a set of incentives andstrategies (Chwe 2001), a typology of behaviors and identity groups(Douglas and Wildavsky 1982), or an enduring product of institu-tional history (Putnam 1993; Rohrschneider 1999; cf. Jackman andMiller 2004).3 Galvan (2004) offers another concept of political culture based inpart on practice, arguing that post-colonial political culture in Sene-gal is a syncretic product of new institutions embedded in traditionalmemory and practice.4 These practices relate to central themes in past studies of politicalculture, specifically, orientation toward authority and collective po-litical conduct (Almond and Verba 1963; Laitin 1986; Weber 1978;Wedeen 1999).

the intricate monarchical system characterized by com-peting patterns of local allegiance into a more central-ized, oppressive, and ethnically homogenous system ofgovernance (Lemarchand 1970; Newbury 1988).

Such examples, however, are small consolation todomestic policy makers and international actors whohope to bring about cultural change. Rarely do thosewho intervene in post-conflict societies have the capac-ity to effect fundamental shifts in political institutions(Jackman and Miller 2004; Putnam 1993) or economicconditions (Inglehart 1997), or the mandate to waitdecades for results. The question of immediate prac-tical concern is whether change can be induced in theshort run. This question is also of enormous theoreticalimportance.

How malleable are the aspects of political culturethought to be associated with genocide? Past researchhas examined cross-regime, cross-national, and long-term temporal variation in aspects of political culture,but thus far no studies have traced change in responseto an experimental intervention. This study consid-ers one such intervention, a Rwandan radio programdesigned to challenge norms of deference, legitimizeexpressions of dissent, and encourage local problemsolving and dispute resolution. We present the resultsof an unusual experiment, one that randomly exposesrural Rwandans either to this program or to a parallelprogram about health over the course of one year. Wemeasured changes in listeners’ attitudes, perceptionsof typical or desirable conduct (i.e., “social norms”;Miller, Monin, and Prentice 2000), and behavior usingan array of qualitative and quantitative measures. Thestudy cannot adjudicate among different accounts ofpolitical culture and cultural change as a whole. How-ever, the results contribute to the study of politicalculture by demonstrating how important aspects ofcultural practice do or do not change in the wake ofexposure to media messages over the course of a year.

Our results indicate that although the program didlittle or nothing to change many domains of individualbelief and attitude, it effected profound changes in be-havior. Radio listeners in the treatment group becamemore likely to express dissent with peers and less likelyto defer to local officials when solving local problems.These changes were balanced by an increased sense ofcollective responsibility and local initiative. Our find-ings suggest that certain aspects of political culture aresusceptible to short-term change in the wake of nonin-stitutional interventions, such as media programs. Evi-dently, the mass media can influence the set of cultur-ally available behavioral practices that citizens use—the“toolkit” of political cultural conduct (Swidler 1986).Using appealing, archetypal rural Rwandan charactersto act out the behaviors of interest, the radio programmade dissent and local collective action acceptable andavailable for use in everyday life. Such changes do notconstitute a fundamental transformation of Rwandanpolitical culture. Still, our results show that modest in-terventions, at least in the short term, can have largeand significant behavioral effects. This represents a sig-nificant step forward for theories attentive to the roleof political conduct in cultural change.

623

Deference, Dissent, and Dispute Resolution November 2009

The article is structured as follows. We begin bydiscussing media interventions as a general tactic forinfluencing culture and the special role of radio in theRwandan setting. Next, we describe in detail the ex-perimental design and measurement strategies used toassess the impact of a post-genocide reconciliation ra-dio program in relation to a comparison program deal-ing with health. The results section describes findingsfrom closed-ended interviews, focus group discussions,role-playing exercises, and unobtrusive measures ofcollective decision making. We conclude by discussingthe implications of our findings for ongoing theoreti-cal debates about media influence, the malleability ofpolitical culture, and post-conflict intervention.

Media as Agents of Change inPost-genocide Societies

In 2004, a nongovernmental organization in Rwandabegan broadcasting a radio soap opera (Musekeweya,or New Dawn) to teach listeners about the roots of vio-lence, the importance of independent thought, and thedangers of excessive deference to authority. This mediaintervention, which enfolds educational messages in adramatic fictional story, is part of a global phenomenoncalled “entertainment education,” which has becomea popular technique among social change campaignsin developed and developing countries (Rosin 2006;Singhal et al. 2004). Such campaigns are especially pop-ular in post-conflict countries where government agen-cies and civil society are often unreliable vehicles forpromoting social or political change. Entertainment-education programs weave messages about health, con-flict resolution, or development strategies into a storyfeaturing realistic and entertaining characters. The the-ory, which draws on psychological social learning the-ory (Bandura 2002), is that media can “both entertainand educate, in order to increase audience members’knowledge about an educational issue, create favorableattitudes, shift social norms, and change overt behav-ior” (Singhal and Rogers 2004, 5).

The soap opera, launched in Rwanda one decadeafter the genocide, featured a fictional story of twoRwandan communities. Due to government restric-tions against public discussion of ethnicity, the storyof the communities serves as a transparent allegoryfor the history of cooperation and conflict betweenRwandan Tutsis and Hutus. Intercommunity tensionscreated by a land shortage are set aflame by dema-gogic authorities who seek to accumulate power, andrelations between the fictional communities disinte-grate as the less prosperous community attacks its ri-val. Against this backdrop of communal violence, aromantic storyline unfolds between a young man andwoman, each from a different community. Instead offalling victim to the violence and prejudices betweentheir two communities, the Rwandan Romeo and Julietform a coalition for peace with citizens from both com-munities. Their coalition defies the power-hungry au-thorities and seeks to mediate the conflict and help thevictims.

We ask the following questions regarding the impactof this media program. First, can a drama written toreflect common experiences of the average Rwandancitizen encourage the adoption of norms and behaviorsportrayed by the program’s heroes? More broadly, canmedia interventions influence an aspect of a country’spolitical culture? Questions about media influence onpolitical behavior and culture abound, but rarely areinvestigators able to observe a media program unfoldand deploy a methodological design that isolates itscausal effects. We conducted a randomized experimentto assess whether one year after the radio programbegan changes could be observed in Rwandan radiolisteners’ deference to authority, willingness to dissent,and collaborative participation in dispute resolution.By deference to authority, we mean submissive be-havior toward another on the part of a subordinateactor or group (Shils 1972), and by dissent, we meanthe expression of opinions that are at variance withthose previously, commonly, or officially expressed bypeers or authorities (Oxford English Dictionary). InKinyarwanda, the language in which the interviewswere conducted, dissent was signified as ndabivuga, “Ispeak [when I disagree],” in opposition to ndicecekera,“I keep quiet.”

The project is part of a broader assessment of theimpact of media on citizen beliefs, norms, and behav-ior. The radio program’s social psychological effectshave been described elsewhere. Paluck (2009) showsthat the radio program shifts listeners’ perceptions ofsocial norms, or perceptions of typical and desirablebehaviors, without affecting factual beliefs regardingthe causes or consequences of those behaviors. In thisarticle, we ask whether the media can influence Rwan-dans’ perceptions of the desirability of certain politicalbehaviors, and whether these perceptions translate intopolitical conduct.

EMPIRICAL METHOD AND RESULTS

Sampling of Participants

We selected 14 research sites to represent salient po-litical, economic, and ethnic categories of present-dayRwanda: two genocide survivor communities (mostlyTutsi), two Twa communities (the Pygmy minority),two prisons, and eight general population communitiesfrom the four general provincial regions of Rwanda.The four provincial regions (excluding the capital) rep-resent differing social and political circumstances. Thenorthwest region was the base of Hutu extremism andsupport for former President Habyarimana, whose as-sassination marked the starting point of the genocide.The current (Tutsi-led) regime distrusts and monitorsthis region. The western region was the site of someof the most rapid genocidal killing in 1994 (in one pre-fecture, more than three fourths of its 59,050 victimswere killed in less than two weeks; Straus 2006, 56–8). The southern region is the historical seat of Tutsiroyalty, and in modern times, has been the intellectualheart of the country, hosting the national university.

624


FIGURE 1. Research Sites by Location and Experimental Assignment

The largest percentage of Rwandan Tutsis lived in thesouth prior to the genocide, and the region had oneof the latest onsets of genocidal violence due to initialresistance by local authorities. The northeast regionwas in the Tutsi rebel (RPF) army’s path as it invadedfrom Uganda in 1993 and 1994, eventually ending thegenocide, but carrying out reprisal violence aimed atHutu along their route to the central capital, Kigali.

Because prison, survivor, and Twa communities arerelatively scarce and scattered across the country, weworked through the prison system and Twa and sur-vivor advocacy groups to obtain demographic, socioe-conomic, and detainee characteristics (for the prisons)for a range of sites, and we chose the two most similarsites of each community type, even if they were locatedin different provinces. We chose each general popula-tion community site on the basis of its accessibility bypaved or dirt road (we eliminated sites unavailable byroad for logistical reasons), and on the quality of itsdemographic and socioeconomic match with anothersite in the region, a site far enough away that the com-

munities did not share markets (to prevent spillover)but close enough to be similar on a range of charac-teristics. We used Rwandan census data to match thesites, including gender ratio, quality of dwellings, re-ligion, and education level. Figure 1 shows the exactlocations of each research site within the four provin-cial boundaries.5

We randomly assigned one site in each pair to listento the reconciliation program, and the other to thehealth program. Random assignment ensures that, inexpectation, the treatment groups have the same ob-servable and unobservable attributes. Block random-ization of sites helps minimize observable differencesbetween the communities in small samples. Even afterblocking, there could be remaining observable differ-ences, but these were controlled in multivariate analy-ses of the outcome data. Because the reconciliation and

5 The four provincial boundaries were drawn after our site selectionin 2004, when the Rwandan government redesigned the country’sadministrative boundaries in 2006 as part of a decentralization effort.

625


TABLE 1. Sample Characteristics by Research Site

General Population Communities

West Northwest Northeast South Twa Survivor Prison

Variable Pairs R H R H R H R H R H R H R H

Sex (% male) 72 58 50 60 56 75 48 33 35 33 68 20 62 69Age (mean) 30 44 37 31 37 42 47 41 37 32 38 44 40 43Lived elsewhere (%) 41 23 75 63 73 60 78 63 50 59 48 44 54 36Attended Primary school (% yes) 77 48 78 80 90 54 63 73 45 54 63 80 67 90Listen radio (%) 92 93 83 80 100 87 75 88 75 85 95 90 74 95Present in 1994 (%) 100 97 100 98 98 100 100 100 100 95 100 100 100 100Lost relative (%) 69 49 33 36 85 70 73 78 59 76 100 100 79 71Relative in prison (%) 51 38 25 3 32 23 38 15 50 65 3 13 46 60n 39 40 40 40 41 40 40 40 40 39 40 41 39 39

R, reconciliation, H, health.Note: Figures are percentages for each research site.

health sites have similar background attributes, thesecontrols turned out to have little effect on the results.

Because radios and batteries are relatively expensivefor Rwandans, they usually listen to the radio in groups.Thus, we used a group-randomized design in whichadults from a community listened together either to thetreatment (reconciliation) program or to the controlprogram (another entertainment-education radio soapopera about health and HIV). Forty adults within eachcommunity were randomly sampled from official listsof the population after stratifying for sex and age (18–30 and 30 and older). The sampling technique invitedat most one person per family.

The first author visited each community with fourRwandan research assistants who represented Hutuand Tutsi ethnic backgrounds. Research assistantssought out the selected individuals and explained thepurpose and procedure of the study. The purpose—“tounderstand Rwandans’ opinions about radio programsand the issues addressed in those programs”—was de-fined broadly so as not to create expectations of anyparticular set of future questions. Nearly all individu-als selected to participate accepted the invitation. Onaverage, one or two selected individuals at each siteturned down the invitation to participate due to poorhealth or indifference. We drew replacements from arandomized backup list created in advance.

Pretest

When an individual agreed to participate, the re-searcher conducted a brief pretest interview to collectinformation about the individual’s background. Base-line pretest data for each community cluster confirmedthat the matched randomization had divided the sam-ple into two groups with similar background attributes.The pretest data also provided covariates that helpminimize disturbance variance in the posttest analyses.

Table 1 presents statistics on the characteristics ofparticipants at each research site. The total participant

sample was 50% male, and the average age was 38.9years. We did not achieve a male–female balance inevery site because of the disproportionate loss of menduring the genocide in some areas (in survivor, Twa,and in two general population villages). Ninety-threepercent of men and 81% of women reported regularlistening to the radio. Only 53% of the participantsactually owned a radio; 83% of those who did nothave a radio reported listening in groups with familyor friends, which affirms the logic of our group listen-ing study design. Seventy-nine percent of participantsidentified themselves as farmers, and 73% of men and67% of women in the total sample had some schooling.The Twa and survivor communities were the poorestof all research sites.

Ninety-nine percent of the sample was in Rwandaat the start of the genocide, and approximately 50%was displaced by the violence for a time of one weekto a few years. Sixty-nine percent of the sample re-ported one or more of their relatives were killed in1994; similar proportions of each community claimedmaterial losses (property looted or burned) during thegenocide. Twenty-eight percent of the general popu-lation participants had a relative in prison comparedto 7% of survivors and 57% of Twa participants. Asa randomization check, we regressed the binarytreatment variable on background characteristics (in-cluding dummies for missing data in the pretest andaccounting for clustered random assignment). An Ftest was insignificant (F = .73, p = .65), meaning thatthere was no unexpected association between charac-teristics of the participants and radio program to whichthey were assigned.

Experimental Design

The same Rwandan research assistants visited eachcommunity over the course of one year and playedeach month’s four 20-minute episodes on a portablestereo for the listener groups. Although research

626


assistants were aware of the program differences, theywere blind to specific research hypotheses. Participantsgathered in their respective community spaces as theydo for nonresearch occasions to listen to the radio. Pro-gram content was the only difference between the twoconditions—listening protocol, frequency, and outcomemeasurements were the same.

Control groups listened to the education-enter-tainment radio soap opera Urunana, which aims tochange beliefs, norms, and behaviors about reproduc-tive health and HIV. Comparing the outcome of groupslistening to a radio soap opera with a different messageisolates the impact of the content of the radio programfrom the social environment that accompanies mediaconsumption. An alternative design would involve ano-listening control group. But the no-listening sub-jects would then differ along many dimensions (e.g.,participation in groups, common knowledge aboutother listeners’ reactions, a monthly schedule) that po-tentially threaten the experiment’s internal validity.

Monthly field visits ensured that participants listenedto the program but preserved the most natural environ-ment possible. Research assistants who visited everymonth sat and listened with the group as part of whatwas designed to be a casual community gathering. Thegroup shared customary local drinks (purchased by theresearch team) and often lingered afterward to chat.Research assistants never initiated or guided discus-sion. They filled out standardized observation sheetsafter they left each site to record attendance and torate the group’s levels of observable enthusiasm, at-tentiveness, confusion, emotional expressions, and theamount and subjects of discussion during and after theprogram.

A content analysis of the radio program’s scripts con-firms the essential similarity of the two soap operas.6As Table 2 makes plain, the principal topics of eachprogram are distinct. Health issues (e.g., AIDS test-ing, health insurance, pregnancy precautions) appearin 83% of Urunana episodes compared to 2% ofMusekeweya episodes. Expressions of dissent, codedwhen opposition to authorities or peers was voiced(e.g., disagreeing with plans to attack the neighbors;denouncing looters; proclaiming friendship with “theenemy”; excluding minor interpersonal disagreement,such as the best way to make banana beer), neverappear in Urunana episodes but arise in 56% of theepisodes in the Musekeweya reconciliation program.That the programs’ messages do not overlap boostsconfidence that any observed change pertaining to thereconciliation program content can be attributed to lis-tening to the reconciliation program. In other respects,the programs are similar in format and present thesame or homologous topics, providing study partici-pants with relatively equivalent listening experiences.Like all soap operas, the radio dramas were filled withromance, humor, villains, deceit, and emotion. In thereconciliation program, the lovers were the Romeo and

6 An independent judge coded a random subsample of the scriptsand reached an acceptable level of interrater reliability with the firstauthor’s codes, α = .91.

TABLE 2. Topics of Reconciliation andHealth Radio Programs as Percentage ofTotal Episodes Played to Participants

Reconciliation Health

Principal topicsDissent 56% 0%Health 2% 83%Topics in commonLove 27% 25%Humor 25% 31%Bad authority 29% 17%Trauma 17% 8%Poverty 19% 13%Mistrust or deceit 13% 23%Trauma healing 13% 8%Homologous topicsPositive authority 19% 2%Hospital 4% 19%

Social cooperation 13% 2%Economic cooperation 2% 10%

Intergroup violence 8% 0%Rape 2% 10%

Empathy 8% 0%Apology 2% 13%Different topicsDiscrimination 17% 2%Marital discord 6% 44%Rumor 21% 6%Intrigue 0% 35%Scapegoat rhetoric 15% 0%Agriculture 2% 10%Friendship 15% 6%

Note: Percentages are calculated by dividing the numberof episodes in which each theme appears by the 48 totalepisodes. Each episode contained more than one theme;thus, percentages do not total 100.

Juliet from opposing communities (27% of episodes);the health program also featured two young lovers, anda comedic and affectionate married couple expectingtheir first baby (25% of episodes). The reconciliationprogram villains were local authorities and elders whofomented hatred and violence between the two com-munities (29% of episodes); health program villainsincluded elders who covered up the rape of a younggirl, an ignorant work supervisor who stigmatized peo-ple with HIV, and “quack” doctors who charged highprices for unhygienic and retrograde medicine (17% ofepisodes). When violence was depicted, it was primar-ily interpersonal (rape) in the health program (10%),and mostly intergroup (attacks and looting) in the rec-onciliation program (8%). Equivalent amounts of con-flict spurred positive but different resolutions in eachprogram: interpersonal apology and economic cooper-ation in the health program (23%), and empathy andcooperation for peace in the reconciliation program(21%). As indicated by Table 2, both programs por-trayed rural Rwanda communities in which some peo-ple lived comfortably but others struggled in poverty.

627


Overall, the reconciliation program focused on so-cial and political conflict and citizens’ responsibility forconflict reduction, whereas the health program focusedon social and familial conflict and citizens’ responsibil-ity for their health. Both programs were penned byRwandan scriptwriters relying on traditional Rwandanstorytelling styles (one reconciliation scriptwriter orig-inally worked for the health program) and were fi-nanced by nongovernmental organizations.

Random assignment of listeners to one of the ra-dio programs addresses the chronic problem of self-selection—listeners’ preexisting tastes may guide theirchoice of media programs and their attitudes and be-haviors. To ensure the health group remained “un-treated” by the reconciliation program, which wasbroadcast nationally during the evaluation period, weasked the health program listeners to refrain fromlistening to the reconciliation program for one year.As an incentive to comply, we promised 14 cas-sette tapes containing the year’s worth of reconcilia-tion program episodes at the end of the year. Thus,health participants understood their promise as a post-ponement and not a sacrifice. To maintain symmetryacross experimental groups, these cassettes were alsopromised to the reconciliation group. (See AppendixI for a detailed discussion of the experimental pro-cedure and the ethical considerations that guided thedesign.)

Health participants do not appear to have listenedto broadcasts of the reconciliation program. We in-structed research assistants to make casual commentsto health participants about the reconciliation programhalfway through the year, but participants indicatedthat they were not listening and seemed unaware of itsmain characters. The high rate of compliance may re-flect the fact that the new reconciliation drama had yetto build up a following, and three alternative programswere aired on other stations in its time slot. If, how-ever, some participants “crossed over” to the treatmentgroup without our knowledge, our statistical findingswould underestimate the true effect of the reconcilia-tion program (Freedman 2006). Our results thereforeprovide a conservative estimate of the reconciliationprogram’s impact.

Data and Statistical Considerations

At the end of one year, a team of seventeen Rwandanresearchers (including those who had visited the com-munities throughout the year) accompanied the firstauthor to each research site for three days. Researchersconducted individual structured interviews, semistruc-tured focus groups, “role plays,” and unobtrusive obser-vations of community negotiations. Before analyzingthe results, we first comment on some statistical issues.

Analytic Issues of Clustered Random Assignment.With fourteen clusters and approximately 40 individu-als per community, the total sample is 556 individuals(fewer than 10 people left the study over the courseof the year; there was no differential attrition for rec-

onciliation and health). However, the “effective N”of this study is lower because individuals within eachcluster are not independent observations. Participantsliving together in each cluster may share unobservablecharacteristics. Their shared characteristics violate theassumption of independent disturbances, an assump-tion on which conventional standard errors are cal-culated. We therefore present robust cluster standarderrors, which account for within-cluster covariance(Arceneaux 2005). We also increase the power of theexperiment by including covariates from the pretest,which potentially reduce the intracluster correlationamong the disturbances and decrease the disturbancevariance.

Evidence of Equivalent Listening Experiences. Amean of 35.6 participants attended each month’s listen-ing sessions at each site, with no meaningful differencein attendance between health (m = 35.8, sd = 3.5) andreconciliation (m = 35.4, sd = 3.7) treatment groups.There were no differences between reconciliation andhealth groups’ observed interest in the program, usingthe research assistants’ ratings combined across all vis-its (using a scale from least to most, 1–5; m = 4.0, sd =.75; m = 4.2, sd = .83, respectively), enthusiasm (m= 3.3, sd = .96; m = 3.6, sd = .90), distraction (m =1.9, sd = .90; m = 1.7, sd = .78), confusion (m = 1.5,sd = .60; m = 1.5, sd = .78), and discussion during thebroadcast (m = 2.9, sd = .90, m = 3.3, sd = 1.1). Afterthe broadcast ended, researchers observed any furtherspontaneous discussion and reported that participantsspent, on average, 63% of their time discussing theprogram (as opposed to other topics) before leaving,a rate that was almost identical in the two experimen-tal conditions. These results indicate that the listeningexperience was similar for reconciliation and health,with the key difference being the content of the mediaprogram.

Individual Interviews and Focus Groups. In thestructured individual interviews, researchers readeach participant a series of statements. Participantsspecified how much they agreed or disagreed with eachstatement by pointing to one of four progressivelylarger circles printed on a large index card; the smallestcircle represented “disagree strongly” and the largest“agree strongly.” We focus here on the item used tomeasure perceptions of a prescriptive (“that is theway people should act”) norm about dissent: WhenI disagree with something that someone says, I shoulddissent. Because the response options are a series ofordered categories, we used ordered probit regressionto analyze the results.

Table 3 shows the effect of the reconciliation pro-gram on responses to this item across three spec-ifications. The first specification includes only therandom treatment (“reconciliation program”) as a pre-dictor. The second includes dummy variables for eachcommunity pair—from the northwest sites to the sur-vivor community sites. The final specification includesthe community dummies as well as measures from

628


TABLE 3. Ordered Probit Estimates ofReconciliation Program’s Effects onWillingness to Express Dissent

1 2 3

Reconciliation .264∗ .285∗∗ .288∗∗

program (.084) (.062) (.058)Northwest −.175 −.126

(.116) (.132)Northeast −.255∗∗ −.189∗

(.076) (.092)South −.177∗ −.105

(.064) (.083)Twa −.305∗ −.252∗

(.109) (.116)Prison −.114 −.085

(.168) (.167)Survivor .047 .111

(.064) (.085)Displaced by −.060

violence (.072)Sex .081

(.109)Age −.003

(.004)Radio listening .071

habits (.082)Cut 1 −.510 −.647 −.655

(.076) (.071) (.228)Cut 2 −.058 −.191 −.206

(.068) (.075) (.227)Cut 3 .270 .138 .122

(.073) (.084) (.213)Log likelihood −626.92 −624.22 −616.03

Notes: ∗p <. 05; ∗∗p <. 001, two tailed. The response scalefor dissent is 1 (I should stay quiet) to 4 (I should dissent);pairs of communities are entered 1-0 dummies for eachgeneral population (NW–S), and Twa, prison, and survivorpair. The omitted pair of communities is the western pair. Sexis a dummy variable with males equal to one. Age is measuredin years. Displacement and radio listening are coded from 1(never displaced, do not listen often) to 3 (displaced a longtime, listen often).

the pretest survey. These pretest measures includedwhether the respondent was ever displaced by the vio-lence in Rwanda, respondent sex, and amount of radiolistening.

The results show that reconciliation listeners were.26 to .29 probits more likely than health listeners toindicate that they should speak up. These estimates arestatistically significant (p < .001) regardless of modelspecification, including the most conservative specifica-tion, a linear model using the research site as the unit ofanalysis (ß = 1.70, se = .63, p < .05, N = 14). They arealso substantively large. A shift of .26 probits implies,for example, that a health group respondent with a 30%chance of strongly agreeing to dissent would move to a40% chance if assigned to the reconciliation programgroup. This is a large, but not implausibly large, shift inopinion.

Focus Group Results

Does a perception that one should express an unpop-ular viewpoint translate into actual behavior? To testthis, we raised the question of trust in both the privateinterview and the public focus group, to determine ifparticipants would express similar views in the two con-texts. From knowledge of the research site and previousresearch (e.g., Gabisirege and Babalola 2001), we ex-pected trust to be low in the wake of the genocide.However, we also knew that the expression of mistrustwould go against the official government line of unityand reconciliation.

In privately conducted individual interviews, partic-ipants were asked to rate their agreement with thestatement: There is mistrust in my community. Thevast majority (80%) of participants agreed with thisstatement. There was no difference in privately re-ported mistrust comparing reconciliation and healthlisteners. This finding is consistent with other evidencedemonstrating that social desirability did not guide theresponses of reconciliation listeners. Paluck (2009) re-ports that reconciliation listeners did not become moresanguine about the effects of intermarriage and othertypes of cross-group social interaction.

Half of all participants were organized into single-sex focus groups of 10 before their private interview,and the other half took part in a focus group after theinterview.7 Researchers posed open-ended questionsand encouraged participants to discuss them with eachother. With little moderation by researchers, the fo-cus groups were more like community forums than-question-and-answer sessions; on average, there were11 contributions for each question, and there was nosignificant difference in the average amount of talkingbetween reconciliation and health focus groups (ß =−4.96, se = 3.59, p = .22).

Among other topics, researchers asked focus groupsto discuss whether there was mistrust in their com-munity. Figure 2 shows that in these public settingsdenials of mistrust became much more frequent—butonly among health groups.8 With the exception ofthe survivor communities (described later), more than60% of the health group comments were denials ofmistrust, compared to less than 20% of reconciliationgroup comments (ß = .96, se = .37, p < .01). In lightof the uniformly high levels of mistrust reported in theindividual interviews, the difference in focus group re-sponses seems to reveal more about the reconciliationgroup’s willingness to speak out on difficult subjectsthan about actual levels of community mistrust.

The one exception to this finding was the survivorgroups—none of whom modified their strong stance

7 We counterbalanced the order of the individual and focus group in-terviews and found no significant differences between the individualswho participated in one sequence or another. We conducted focusgroups and role plays in only eight of the 14 sites due to logisticalconstraints.8 Going beyond the data depicted in Figure 2, we note further that39% of the health groups’ comments were unqualified denials of anymistrust in their community (e.g., “there is no mistrust at all!” asopposed to statements downplaying mistrust), compared to 7% inthe reconciliation groups.

629


FIGURE 2. Denials of Mistrust Are More Frequent When Made in Public among Health Listeners

Focus groups were only conducted in prisons, survivor communities, and south and northwest communities (combined here into “generalpopulation” communities).

that there is mistrust in their community. A plausibleexplanation for this break in the pattern, which wediscuss further in this section, is that their identity asgenocide survivors and as coethnics with the currentregime grants them greater license to speak out aboutthe social effects of the genocide. Although the cur-rent governmental regime pressures Rwandans to ad-here to their official line of “unity and reconciliation,”genocide survivors (backed by strong political interestgroups within Rwanda and featured in government ap-peals for aid) are to a certain extent granted exception.

Social Distance Attitudes

It should be emphasized that the behavioral changesjust described were not accompanied by a more generalwillingness to affiliate with members of other groups.Our questionnaire asked whether participants were in-terested in greeting, sharing beer with, working with,and intermarrying a child with a member of a different

group (the implication was a different ethnic group,although the prohibition on using ethnic terms pre-vented us from specifying this directly).9 We found noreduction in social distance when comparing the healthand reconciliation groups. The null results reported inTable 4 are important insofar as they run counter to thehypothesis that demand characteristics of the experi-ment caused respondents in the reconciliation group toexpress more tolerant opinions.10

9 See Appendix II for the full questionnaire administered to thereconciliation and health groups.10 The fact that some survey outcomes show significant effects,whereas others do not, raises the issue of how to maintain the appro-priate test size across multiple comparisons. The questionnaire (seeAppendix II) contains four distinct dependent variables: an indexof questions concerning social norms, an index concerning factualbeliefs, an empathy scale, and a social distance scale (see Paluck2009 for a list of the factual and social norm indexes). The ques-tionnaire contains four additional items related to program content,which we do not examine here. A Bonferroni correction for the

630


TABLE 4. Linear Regression Estimates ofReconciliation Program’s Effects on SocialDistance

1 2 3

Reconciliation program −.029 −.029 −.041(.135) (.086) (.085)

Northwest .050 .018(.097) (.068)

Northeast .285∗ .287∗

(.114) (.083)South .158∗ .157

(.103) (.076)Twa −.069 −.079

(.271) (.250)Prison .241∗ .237

(.130) (.117)Survivor −.284∗ −.317∗

(.114) (.094)Displaced by violence .006

(.047)Sex .068

(.092)Age −.003

(.004)Radio listening habits .110∗

(.045)Constant 3.351 3.300 3.215

(.087) (.104) (.152)

Notes: ∗p < .05; ∗∗p < .001, two tailed. Social distance isan index of willingness to greet, share a beer, work together,or marry child to another group. The response scale for eachsocial distance item is 1 (not willing) to 4 (very willing). Pairsof communities are entered as 1-0 dummies for each generalpopulation (NW–S), and Twa, prison, and survivor pair. Theomitted pair of communities is the western pair. Sex is adummy variable with males equal to one. Age is measuredin years; displacement and radio listening are coded from 1(never displaced, do not listen often) to 3 (displaced a longtime, listen often).

Role Playing

We were also interested in measuring group-generatedoutcomes because dissent and deference to authority isexpressed socially and conditioned by one’s social sur-roundings. In particular, we want to test participants’approach to a collective dispute. To single-sex groups ofparticipants at each site, we played recordings of unfin-ished radio scenes written by the nongovernmental or-ganization’s (NGO’s), scriptwriters and the first author.Each unfinished scene centered on a common problemin Rwandan society that had also been addressed in thereconciliation radio program. In the scene relevant tothe present analysis, the characters argue about how tohandle a group of refugees fleeing a famine.

This refugee scene (reprinted in the Appendix III)presented three main characters. One alarmed and an-

target significance level of .05 would therefore be .05/8 = .0063. Theestimated treatment effects in Table 3 remain significant despite thisquite conservative threshold.

gry man comes running to tell two neighbors that hehas learned about the fleeing neighboring community.He exclaims that their own community is too poor tohelp the refugees and asks the other two to help himblock the border and, if necessary, use force to repulsethe refugees. The two neighbors hearing the news aredivided: one agrees, but the other argues that theyshould welcome the refugees. The three men debate,raise a few different points about why they should orshould not help the refugees (i.e., violence is wrong,these neighbors may help us in the future, we have nomoney, I suspect them of stealing money from me inthe past), when the scene suddenly ends. The narratorwho introduces the scene announces that she wouldlike the participants to act out the end of the scene:“What do you think should happen now?”

This exercise was performative by definition—ouraim was not to approximate what group memberswould actually do in such a situation, but rather togauge their prescriptive ideas about what they shoulddo. Educators, therapists, and activists use participatorytheater in other settings (e.g., Boal 1992; Breed 2007),but rarely have social scientists measured group dy-namics or social prescriptions with role playing, makingthis an exploratory measure. After listening to a sceneabout a community meeting that familiarized themwith the exercise, both male and female participantsseemed to enjoy and even “lose themselves” in therole plays. One group of women staged an argumentin their practice scene with such conviction that thegroup attracted the attention of a military guard whowas patrolling the general area (he was persuaded toleave by the first author).

Seventeen groups were formed across eight researchsites. Each group, on average, performed approxi-mately two unique new endings (allowing differentmembers of the group to assume the primary roles) fora total of twenty-one scenes. Typically, the entire groupparticipated in each scene; those who did not assumeone of the three primary roles acted as bystanders orinvented new characters who joined the action (e.g., alocal authority or a family relation of a main character).Two researchers hand-recorded their speech and actionin script format.11

Content Analysis of Role-Play Scripts

The first author coded each speaking turn in the“scripts” that resulted.12 We were interested in howeach group solved the problem (i.e., by welcomingthem or mobilizing in various ways to provide forthem). We noted patterns in the vocabulary partici-pants used to argue for and against various ideas.

Problem Resolution Codes. We coded whether thescript ended by helping the refugees; if the refugees

11 Hand-recording the scenes meant that we captured a less thanfine-grained account of each scene, but this was superior to a taperecorder for its relative unobtrusiveness; we were concerned partic-ipants would feel inhibited with a mechanical recorder.12 An independent judge who coded a random subsample of thescripts reached an acceptable degree of interrater reliability, α = .89.

631


were offered help, we coded how participants mobi-lized to help them. It turned out that the means ofhelping, and not whether they did help, was the dis-tinguishing factor among the scenes. Specifically, wefound, as illustrated in the following two excerpts, thatactors either helped the refugees themselves or calledon the Rwandan government or on NGOs or interna-tional organizations (IOs) for help.

Scene resolution 1: Helping refugees with own resources

[Action: The refugees, dragging their feet and lookinghungry and tired, approach the two men at the borderwho want to fight them (Mukama and Cacana).]

Mukama: Good morning. (pauses) How are you?Refugee A: Nothing goes right. We are hungry and

tired, and we seek your generosity. Ourlife is in your hands.

[Action: Mukama and Cacana walk away from therefugees to talk, where others from their communitystand watching.]

Cacana: Hmm, Mukama. Do you know what?One never knows. What if we don’tsave these people, in the future wemay need their help. Would they helpus? Look at them, and imagine theyare us. I think Kandonyi [3rdcharacter advocating for aid] wasright. Let’s welcome them.

Mukama:(turns to all)

You are welcome in our community.We will share what we have, even if itis not enough. Aho umwaga utariuruhu rw’urukwavu rwisasira batanu.(Where there is no greed, five peoplesleep on one pallet made of a rabbit’sskin.)

Refugee B: I see that people in need make youfeel pity, and empathy. These othermen Mukama and Cacana saw themand were taken by pity. I think weshould thank and applaud thedecision they took.

[END]

Scene resolution 2: Helping refugees with state or NGOresources

Bystanders: They are also human! Why stopthem? Why shouldn’t we share?

Cacana: (tobystanders)

Are you going to make the rain?

Bystander: The HCR (United Nations HighCommission for Refugees) shouldintervene!

Cacana:(to firstbystander)

Are you are going to put them inyour family’s house?

Local mayor: Let them come, we are going to lookfor the donors.

(Later)

[Action: Important local authority calls a meeting withthe ten smaller authorities (Nyumbakumi), each incharge of a block of ten households.]

Localauthority:

Here’s the problem: we have thesepeople who are coming. I want tohear people’s point of view. Cacana,tell us your point of view.

Cacana: They are coming to invade us! Theycould have ulterior motives. Theystole money from Mukama, they mayhave other reasons for coming.

Localauthority:

Everyone should speak.

Executivesecretary:

Cacana shouldn’t prevent anyonefrom coming. There are the NGOswho provide for refugees.

Nyumbakumi1:

Cacana has a very bad idea. Let thepeople come.

Nyumbakumi2:

God is there, he should intervene,Cacana should be punished.

Nyumbakumi3:

Bring the refugees to the sector officeso they can be accessible to theNGOs and donors.

Localauthority:

As the population is for this, therefugees should come.

(Later)

[Action: Approval of the group, arrival of donors.]

Language. We also content analyzed the vocabularyused in each scene. A remarkable aspect of these tran-scripts is the way the “public transcript” of the Rwan-dan government appeared to influence the languageused in each research site. Following Scott (1990),we use public transcript to mean government dis-course, with which Rwandan peasants are familiarizedthrough speeches broadcast on the radio and in com-munity “sensitization” meetings in which local author-ities teach citizens about new policies.

Some of the most important policies publicized andenacted by the Rwandan government in recent yearsdraw on language of forgiveness (e.g., community courtprocesses that emphasize public apologies and acts offorgiveness), reeducation (e.g., ingando or reeducationcamps to which newly released prisoners convictedof crimes of genocide visit before returning home),and decentralization of state power. This discourse ap-peared frequently in the role plays, for example: “Ipublicly ask for forgiveness for trying to block therefugees”; “These people have bad ideology and needto change!”; and “As the decentralization system stip-ulates, it’s up to you to take a final decision.” We codedeach of these speaking turns to note their themes andprevalence.

Role-Play Results

We coded 359 “lines” from the scripts, which includeturns of speech and notes about action. Each scenefeatured, on average, seventeen unique speech or ac-tion lines, and each line could receive more than one

632


FIGURE 3. Content Analysis of Reconciliation and Health Listeners’ Resolutions ofRole-Play Dilemma

Percentages were calculated by dividing the number of scenes that were resolved in a given way by the total number of scenes enacted,across all communities in the Health and Reconciliation conditions.

code.13 We calculated the relative frequency of eachcode in terms of the total number of codes using twomethods: weighting all sites equally, and weighting ac-cording to scene length. The results do not changeappreciably in either method (see Appendix IV forcalculations), and we present results weighting sitesequally in Figure 3.

Problem Resolution Results. In every role play, par-ticipants decided to welcome the refugees and to shameand sometimes punish those who wanted to keep them

13 There were significant differences among the seven researchers interms of number of spoken lines recorded: four hand-recorded anaverage of 14 lines per scene, and three others averaged 8. Whenthe number of lines is smaller, the length of each line is longer; thus,it appears scribes made trade-offs between fidelity and quantity ofcomments recorded. Because both longer and shorter scribes wereassigned to both reconciliation and health groups, this does not biasthe results.

out. The striking difference between reconciliation andhealth scenes lies in how participants arranged to carefor the refugees. The two dominant motifs were to han-dle the problem within the community, by collectivelyorganizing shelter and gathering resources from eachfamily, or to request assistance from the governmentor NGOs/IOs.

Figure 3 shows that reconciliation groups neverasked for help from the government or from an NGO/IO—each time they decided to welcome the refugees,they agreed that they would share with them “the littlethey had.” By comparison, 16 of the health group’s 21“resolution” comments about how they would providefor the refugees involved asking the government or anNGO to help out (p < .001). Requesting assistancefrom the state or aid organizations was not a domi-nant theme of the health program (see Table A1 inAppendix V for precise ratios of speaking turns fo-cusing on community vs. government/NGO/IO help).

633


Some reconciliation scenes did not reach one particu-lar conclusion about how to provide for the refugees(e.g., in three scenes played at the reconciliation prisongroup, one role-play group self-organized to help therefugees, and the two others were resolved by punish-ing those who wanted to fight off the refugees). In thehealth condition, some scenes chose both communityand government/NGO/IO help (e.g., deciding to feedthe refugees until the Red Cross came).

Despite this striking difference between local re-sponsibility and external aid, it is important to pointout that both reconciliation and health participants in-vented authority characters for each scene. Twenty-six percent and 15% of all “process” comments inscenes in health and reconciliation groups, respectively,concerned the local government. Characters in scenesacross all research sites called on important local au-thorities or on the police for help and advice.

Results for Language Use. The percentage of com-ments that used government language shows that, onaverage, 8% of all comments made in each scene drewfrom the “public transcript.” There were no differencesbetween the reconciliation and health listeners in thepercentage of comments using government language.More important, the use of government language wasnot reliably associated with any particular decision re-garding the refugee dilemma. Participants used govern-ment language to justify whatever actions they took.(“This is the time of decentralization” was used to ar-gue for providing for the refugees themselves and alsofor applying to international organizations such as theRed Cross for help.) This finding reveals participants’opportunistic use of elite parlance, rather than a passiveacceptance of media-borne messages “coming from thetop.”

Behavior: Unobtrusive Measurementof Community Negotiations

Thus far, we have measured behavior within artificialsettings—in focus group discussions and in role-playingactivities. We also sought to trigger a more naturaldilemma for our participants that would not feel simu-lated or monitored so that we could observe behaviorthat is closer to “actual” behavior. At the end of ourtime in each community, the first author and the re-search team gathered community members togetherto thank them for their participation and to share re-freshments. During this goodbye party, we presentedthe group with one battery-operated portable stereoand a set of 14 cassette tapes of the radio program. Weacknowledged that there was only one, and suggestedthat because they were all together they could use thistime to decide how to share the stereo, when the groupmembers would listen, and how to keep it suppliedwith batteries. After this casual suggestion, the researchteam thanked the participants and sat back down.

Given the monetary value of a portable stereo, thisdiscussion was of great significance to the participants.The measure also captured spontaneous behavior thatparticipants believed to be “off record”—their discus-

sions took place immediately after the research team’spresentation. Two researchers sitting discreetly in theback of the group recorded these ensuing discussionsby hand, with notebooks in front of them to appear asif they were finishing notes from earlier in the day.

Did the reconciliation radio program’s messageshave an effect on actual participant behaviors regard-ing community issues of consequence? Of interest waswhether the participants allowed for multiple view-points and whether they expressed optimism aboutcooperating to share this community property. We werealso interested in the groups’ ultimate decisions abouthow to share the communal stereos.

During the health groups’ deliberations about theportable stereo and cassettes, we frequently observedthe following pattern: the first member of the groupto speak would propose handing over the stereo andcassettes to the village’s local authority, who couldregulate usage and financial contributions for the bat-teries. Following this proposal, group members wouldoverwhelmingly support the motion and close the dis-cussion.

In the reconciliation groups, deliberations typicallyfollowed a different pattern. After the same initialproposal to entrust the stereo to the authorities, oneor more of the participants would challenge this sug-gestion, for example, claiming that the group shouldbe collectively responsible or should elect one of theirmembers to manage it. Comments about the group’sability to cooperate came up more frequently, such as“We’ve been coming together to listen all of this time,why can’t we come together to listen to this stereotogether, just as we did before?”

These different patterns were borne out by statisti-cal analyses of the coded transcripts. Table 5 presentsthe number of dissenting propositions that followedthe initial proposal to assess the extent and opennessof the deliberation session. This indicator reveals thatreconciliation groups proposed and debated a signif-icantly greater number of views on how to share thecommunal stereo compared to health groups (using aMann-Whitney test, z = −2.6, p < .01; using Wilcoxonsigned-rank test, which controls for the site strata, z =−2.2, p < .05). The number of positive comments madeabout group cooperation was also more frequent inreconciliation groups (using a Mann-Whitney test, z =−2.6, p < .01; using Wilcoxon signed-rank test, whichcontrols for the site strata, z = −2.2, p < .05). Im-portantly, these results were the same for ethnicallyhomogenous Twa and survivor communities and forheterogeneous communities in the general population.

Heterogeneity of Experimental Effects

Table 6 summarizes individual- and group-level resultsfor each separate research site, underscoring the re-markable homogeneity of the reconciliation programeffects across a range of Rwandan communities. Asdescribed in this section, we do find exceptions tothis homogeneity in two ethnic subcommunities: thenorthwest (predominantly Hutu) communities, and the

634


TABLE 5. Behaviors Observed During Discussion of Community Resource

Pair Health Reconciliation

Number of dissenting opinions General population a 1 1b 0 0c 0 2d 0 1

Twa e 0 2Survivor f 0 1Prison g 0 2

Mann-Whitney test z = 2.64, p < .01Wilcoxon signed-rank test z = 2.17, p < .03

Number of cooperative comments General population a 0 2b 0 1c 0 2d 1 1

Twa e 0 0Survivor f 0 1Prison g 0 1

Mann-Whitney test z = 2.59, p < .01Wilcoxon signed-rank test z = 2.17, p < .03

Notes: Letters a–g each represent a pair of communities; in each pair, one community was randomlyassigned to listen to the health program and the other, the reconciliation program. Transcripts from communitydiscussions were assigned one point for offering a dissenting opinion about the initial position taken byone or more members of the group and for any comment made about the group’s ability to cooperate.

TABLE 6. Overview of Results

1 2 3 4 5 6 7

I Should Role Play: Role Play:Pair/ Expt. Stay Social Deny Reliance on Reliance on Dissent in Cooperate inRegion Group Quiet Distance Mistrust Local Aid Government Aid Negotiation Negotiation

(−) (+) (−) (+) (−) (+) (−)

West R 2.03 3.40 1 2West H 2.14 3.18 1 0NW R 2.27 3.32 1 4 0 0 1NW H 2.29 3.35 5 1 0 0 0NE R 2.31 3.64 2 2NE H 2.44 3.51 0 0South R 2.16 3.38 0 5 0 1 1South H 2.41 3.50 2 2 6 0 1Twa R 2.11 2.88 2 0Twa H 2.70 3.60 0 0Prison R 1.70 3.68 0 1 4 2 1Prison H 2.57 3.45 2 1 0 0 0Survivor R 1.90 3.09 0 5 4 1 1Survivor H 2.19 2.95 0 0 0 0 0Experimental .29∗∗ −.04 .96∗ 3.0∗ −3.5∗ 2.6∗ 2.6∗

differencea

R, reconciliation; H, health.Notes: ∗p < .05; ∗∗p < .01. Pluses and minuses represent the prediction for each variable, where a plus indicates that we expecthigher scores for the reconciliation listeners.Entries for columns 1 & 2 are averages of the ratings reported by each individual to statements about dissent and social distance;entries for columns 3–7 are counts of group-generated behavior observed in focus groups (3), role plays (4 & 5), and communitynegotiation (6 & 7).a Columns 1 & 2 present ordered probit estimates of treatment effects at the individual level, columns 3–5 present ordered probitestimates of treatment effects at the group level, and the Z statistics in columns 6 & 7 are Mann-Whitney test results of treatmenteffects at the group level. Focus groups and role plays were not conducted in Twa, western, and northeastern pairs.

635


genocide survivor (Tutsi) communities. We test but findlittle evidence of significant interactions between indi-vidual characteristics and exposure to the reconcilia-tion program.

Community Characteristics. At the communitylevel, Hutu living in the northwestern research sitesand Tutsi living in the genocide survivor communi-ties deviated from the general pattern of experimentalfindings. Although our small sample of each type ofcommunity does not allow us to rule out the possibilityof chance fluctuations, these interactions between com-munity identity and reaction to the radio program seemto accord with the communities’ political standings inRwanda.

As described previously, the predominantly Hutucommunities in the northwest region (the seat of Hutuextremism prior to the genocide) are closely moni-tored by the current Tutsi regime. During data col-lection, citizens in this region were keenly aware ofgovernment surveillance as a result of attacks launchedby ex-Habyarimana military from the nearby border,which were reportedly supported by members of thelocal population (in focus groups, many participantsdiscussed the importance of trusting that your neighborwill not harbor “infiltrators” and thus invite repercus-sions on the whole community). Table 6 shows thatunlike the rest of the sample, the northwest reconcil-iation listeners were not more likely than the north-west health listeners to perceive a norm of dissent oractually to dissent in the group deliberation (in fact,neither group dissented at all). However, we observean experimental effect on northwest reconciliation lis-teners’ open discussions of mistrust, cooperation ingroup deliberation, and reliance on local aid in the roleplays. These findings tell a story consistent with theregion’s political standing: the reconciliation programsuccessfully facilitated better local interaction and co-operation, but had no effect on willingness to speak outin the context of government surveillance and censure.

Results for the genocide survivor communities standout in a different way. Genocide survivors are the leastlikely of all communities in the sample to be cen-sured for politically incorrect speech, as coethnics ofthe current government and as the government’s livingjustification for speech restrictions and internationalaid. At odds with government rhetoric about Rwan-dan unity, and contrary to the rest of the communities’patterns, no genocide survivor publicly denied the exis-tence of mistrust in their communities. In addition, Ta-ble 6 shows that reconciliation listeners in the survivorcommunity called on the government for aid almost asfrequently as they offered their own aid in role plays.In line with the rest of the sample, however, survivorreconciliation listeners were more likely to perceive anorm of dissent and to express dissent in the groupdeliberation compared to survivor health listeners.

Although the two communities react differently tothe radio program with respect to certain outcomes,it is important to remember that these communitiesare similar to other communities with respect to otheroutcomes—dissent among survivors, and collective ac-

tion among the northwest communities. In other words,the communities’ politicized identities did not pushthem entirely out of the radio’s range of influence.

Individual Characteristics. Our data do not revealdifferences in experimental effects according to indi-vidual characteristics such as age, experience of vio-lence, and access to media (see Appendix V, TablesA2 and A3). Although some theories of political cul-tural change predict that younger generations are morelikely to adopt new cultural values (Inglehart 1997, 15;c.f. Prewitt 1975), like Rogowski (1974), we find no realdifferences for older and younger cohorts.14 Models ofself-interest (Sears and Funk 1991) or grievance andblame (Gurr 1970; Iyengar 1991) might predict theradio impact would vary according to the magnitudeof participants’ personal losses in the 1994 violence—but we do not find this to be the case. Treatment effectsdo not appear to vary by a pretest question askinghow many family members were killed by war, geno-cide, or related violence in 1994. Finally, theories ofmedia sophistication and socialization (Gerbner et al.2002; Heath and Bryant 1992) would predict that radioownership might affect participants’ responses; radioownership also does not interact with experimentalmedia exposure. This finding also addresses concernsthat participants with radios surreptitiously listened toradio programs not assigned to their group in their owntime. If contamination had been a serious problem, wemight have found different effects for those who ownradios.

Discussion

Our evidence suggests that certain behavioral aspectsof political culture are malleable in the short run. Amass media program was sufficient to shift perceivednorms of open expression and local responsibility forcommunity problems, as well as actual open expressionand dissent about sensitive community issues such astrust and resource distribution. Listeners did not be-come contrarians or antiauthoritarians—for example,when reconciliation listeners decided to take collectiveresponsibility for the hypothetical refugee problem,they created roles for local authorities in their decision-making process. Moreover, the radio program was nota panacea—attitudes toward interaction across sociallines were resistant to change. Personal convictionsabout social group boundaries may be more difficultto change; however, as we have shown in other work,personal convictions may be less important motiva-tions for political behavior than perceptions of what issocially and politically normative (Paluck 2009).

14 We used age cohorts that represent significant periods of Rwandanhistory (a child cohort age 12 or younger in the 1994 war and genocideand 18–24 during the study, young adult cohort age 13–25 in 1994,an older adult cohort age 26–41 in 1994 that came of age after in-dependence and the “Hutu revolution” overthrowing Tutsi rule, andan elder cohort age 42–75 in 1994 that grew up before independenceand the Hutu revolution. The lack of apparent interaction effects isinteresting given that the middle cohorts were the most likely to beinvolved in the violence of 1994 (Verwimp 2005).

636


When accounting for change or continuity in someaspect of political culture, most theories point to far-reaching changes in institutional structure or economicconditions (Eckstein 1988; Inglehart 1997; Laitin 1986).We do not claim that the radio program, after oneyear, induced transformative change in Rwanda’s po-litical culture. However, the radio drama’s portrayal ofRwandan attitudes and behavior toward authority andtoward other citizens appears to have influenced theavailability of a certain set of behavioral practices (orthe cultural “toolkit” of behaviors; Swidler 1986). The-ories of political culture that take seriously the role ofpolitical practice (see Sewell 2005) might regard this asa potential first step toward more fundamental culturalchange. We return to this point later.

The behavioral changes induced by this radio pro-gram seem to reflect something other than “blind obe-dience” to the radio or to the research team. For one, weobtained our key measures of behavior in a situationthat was putatively off record, in which participantsbelieved researchers were not gauging their response.In addition, we repeatedly find that listeners did notmerely endorse what they heard on the radio—recon-ciliation listeners were not more likely to report thatthey would affiliate with members of another group, didnot adopt beliefs the soap opera encouraged (reportedin Paluck 2009), and selectively aligned with charactersthat broke with rather than conformed to traditionalexpectations.15 As further evidence against passive ac-ceptance of media messages, role-play participants inboth listening groups opportunistically used the samegovernment rhetoric to justify different decisions abouthow to resolve the hypothetical refugee crisis.

The direction of media influence in this study is, inlarge part, vertical and top-down. It is a program writ-ten by educated Rwandans and broadcast on nationalradio. However, our results suggest that the radio pro-gram sparked horizontal peer influences as well. Re-search assistants’ observation notes from the listeningsessions describe high levels of spontaneous discussionabout both media programs. The participants’ individ-ual and group-generated responses (regarding socialnorms, speaking about sensitive social topics, respond-ing as a group to a hypothetical crisis, and group man-agement of a communal resource) suggest that they hadcome to a joint understanding about the acceptabilityof dissent and desirability of collective action in theirgroup. Certain group responses may also have beenmoderated by assessments of the local political context(e.g., northwestern Hutu communities). Separate re-search programs in very different contexts concur thatmuch of the media’s influence is channeled by audiencediscussion (e.g., Katz and Lazarsfeld 1955). We suggest

15 The content analysis shows that the reconciliation program pre-sented both dissenters and obedient followers, good and bad author-ities. One interesting fact is that the program’s primary villain, theleader of the perpetrator community, was played by the best knownand loved actor in Rwanda. Although this might have afforded hischaracter a more sympathetic hearing among listeners, this does notseem to be the case.

an appropriate way to think about the media’s role ininfluencing political culture is as a connective tissuethat transmits messages about culturally appropriatebehavior, messages that are discussed and, in somecases, adopted by all levels of society.

A balance of dissent and collective action emergesfrom various measures. We found an increase in theperceived acceptability of dissent, as well as actual dis-sent in the individual interviews and the focus groups.Then, holding dissent constant (by presenting dissent-ing actors in the unfinished radio scenes to all par-ticipants), we observed an increase in willingness tocollectively provide for hungry refugees among thereconciliation listeners. Our behavioral measurementstrategy captured the two phenomena at once: we ob-served more dissenting opinions on how to share thecommunal resource (the stereo and cassettes), alongwith more expressions of cooperation about how toaccomplish this feat among reconciliation listeners.

From a normative standpoint, a balance of dissentand collective action seems desirable. On one hand,fully realized collective action with little dissent maylead to phenomena such as blind obedience, crowdbehavior, and collective violence. As Gourevitch (1998,95) observed, “[g]enocide, after all, is an exercise incommunity building.” On the other hand, unrelentingdissent thwarts collective action. This dilemma callsfor a behavioral theory of dissent and collective actionthat describes mechanisms for achieving productivebalances between the two. One mechanism used bysocial psychologists to explain collective action is col-lective efficacy, or group members’ belief in the powerof the group to achieve goals through collective action(Bandura 2000, 75). Researchers propose that the act ofparticipating in group discussion and decision makingenhances the sense of efficacy necessary to the exe-cution of collective action and may teach individualsto avoid blind obedience (e.g., Kelman and Hamilton1989, 324–5).

Political science is concerned with the ways insti-tutions are internalized—it is one thing to write leg-islation and enact policies, and it is another for citi-zens to recognize opportunities and constraints set bythese institutions. Clearly, mass communication is animportant way in which institutions are translated forpublic understanding.16 This leads to the question: isinstitutional change a necessary condition for changingpublic understanding of institutional opportunities andconstraints? The provocative suggestion of our currentfindings is that perhaps it is sufficient to encouragea new understanding of social norms of dissent, defer-ence, or dispute resolution using the media, rather thanto launch far-reaching changes in law or policy. Percep-tions of social norms—of the typicality and desirabilityof certain behaviors—may be sufficient to alter whichbehaviors are selected from the “cultural toolkit.”

16 Wilson’s (2000) review of the political culture literature discussesthe importance of transmission mechanisms for public understand-ing, but surprisingly leaves out the mass media.

637


Restated, perhaps it is possible to instigate changesin deference, dissent, and collective action by modelingor communicating new social norms rather than by re-shaping institutions (formal or informal). This study didnot track the effects of an institutional change, yet theobserved shifts in rural citizens’ normative perceptionsand behavior were of the magnitude that we mightexpect from institutional or economic adjustment(Jackman and Miller 2004), not from a radio soapopera. Most citizens have very poor understanding ofhow institutions work, even those especially interestedor motivated (Robinson and Darley 2003). Citizensalso refashion externally imposed institutions accord-ing to local traditions (Galvan 2004). Our propositionis social psychological as opposed to structural: whatmatters, most social psychologists would posit, is theway people perceive rules and structures, rather thanthe rules or structures themselves. Incentives and con-straints can have psychological reality even without astructural basis. Institutional changes and noninstitu-tional changes may be close cousins when they affectcitizens’ perceptions of descriptive norms (i.e., percep-tions of what others in their environment are doing)and prescriptive norms (i.e., perceptions of what peoplein that environment should do).

These results suggest that new perceptions of normsand new patterns of political conduct can develop with-out institutions, but the next question is whether theycan endure without institutional backing. This ques-tion is of special relevance for Rwanda, where thecurrent government relies on the same mechanismspreviously recognized as facilitating conditions for sub-ordination and conformity. The government has largelymaintained a pre-1994 bureaucratic structure and reg-ulation of rural life, including “sensitization” meetingsin which peasants gather to receive instructions fromstate-designated local authorities and umuganda, orforced monthly labor, for the state. Speech is limited—citing previous hate media, the government censorsjournalists and only recently allowed private radio andpress. Speech about ethnicity is criminalized as “di-visionist.” One wonders whether a fledgling norm ofopen dissent can long survive in this context. More-over, our own findings remind us of the coercive powerof institutions for shaping political behavior. The dif-ferent patterns of dissent we found in communitiesthat are favored or mistrusted by Rwanda’s currentregime are consistent with other micro-level researchshowing how institutional authority can channel mediamessages in Rwanda, with important behavioral con-sequences (Straus 2007).

Given the behavioral constraints that Rwandansface, it is unclear whether the behavioral changes in-duced by the reconciliation program can precipitatebroader changes in Rwandan political culture. We lacksystematic evidence with which to gauge long-termchange in Rwandan political culture. Recall that ourexperiment concluded after one year, at which pointwe provided the reconciliation program to the healthgroup, and follow-up research showed the expectedconvergence in survey responses across the two exper-imental groups a year later. To assess the long-term

impact of the reconciliation program, now into its fifthyear of broadcast, we would at very least need longitu-dinal survey data, but to our knowledge no such surveysexist. From our long-term research engagement in thecountry, we know that knowledge of the soap opera iswidespread, but its long-term effects remain unknown.Research in years to come may investigate the rateof change and the dynamics of contestation betweentradition and transformation in Rwanda.

Deepening our understanding of media influence,behavioral change and political culture requires morestudies in a variety of media and institutional settings.This study presents a provocative set of findings, butmore important it provides a template for future work.This template brings together a number of elementsthat are often recommended but seldom used in con-cert: the use of random assignment to create treatmentand control groups, the study of real world media in-terventions in naturalistic settings, and the develop-ment and implementation of a variety of qualitativeand quantitative approaches to the measurement ofoutcomes. Like other recent experiments and quasi-experiments in developing countries (Blattman 2008;Wantchekon 2003), this study calls into question thepresumed infeasibility of this type of political scienceresearch. Nevertheless, this line of research remains inits infancy, and the task ahead is to use this template tostudy longer time frames, different interventions, andvarying political contexts.

APPENDIX I: PROCEDURAL AND ETHICALPROTOCOL DETAILS OF THE EXPERIMENT

To conduct this experiment, we worked in close collaborationwith the nongovernmental organization (NGO), LaBenevo-lencija, a Dutch-based NGO that works in the Great Lakesregion of Africa (specifically, Rwanda, DRC, and Burundi).The NGO began work in Rwanda in 2003, with funding frominternational government donors. The direct research costsof this study were funded by budget lines the NGO’s donorsreserved for an evaluation, as well as by funds the authorsraised in academic grants awarded specifically for this projectand in funds for field experimentation conducted out of theInstitution for Social and Policy Studies at Yale.

Our collaboration with the NGO began when the firstauthor contacted the NGO in 2003, after learning througha Web site about its aims to create a radio soap opera. TheNGO’s stated purpose, in their words, was to create more“active bystanders” in the population, to encourage citizensto speak out against discrimination and violence, and to ed-ucate people about the evolution of violence. The NGO hadsolicited the help of two psychologists to provide the Rwan-dan scriptwriters with theoretically informed messages aboutdissent and the progression of violence.

The first author spent one month conducting qualitativeresearch in Rwanda for the NGO, directing focus groups withrural Rwandans about media and the topics of the prospec-tive soap opera. After this research, the authors presentedthe NGO with a methodological design for measuring theimpact of the radio soap opera. The NGO contracted usto conduct their impact evaluation as pro bono consultants;their two stipulations were that the study test acceptance ofspecific messages written by the consulting psychologists, andthat it sample from the Twa, genocide survivor, and prison

638


subpopulations, as well as the general population (to reflecttheir wide target population as well as the sample used inbackground research on the soap opera plot conducted bySuzanne Fisher and Solange Ayanone; see Fisher 2004).

We sought ethics reviews for the proposed study protocolin Rwanda and in the United States from Yale’s institutionalreview board (IRB). Because there was no official IRB inRwanda at the time, we reviewed our design with severalRwandan academics, with international researchers who hadexperience conducting surveys or program evaluations inRwanda, and with advocacy and human rights groups inRwanda, including prison advocacy groups. Yale University’sIRB reviewed and approved the entire research design, in-cluding the unobtrusive behavioral measurement.

The final research protocol, developed under the super-vision of the Yale IRB, included a protocol for participantinformed consent. This verbal consent informed participantsthat they could opt to drop out of the study at any time orrefuse to answer any questions posed to them. It warned themabout the possibility that some questions might be emotion-ally evocative and underscored their right to free psychologi-cal treatment for emotional disturbance. The Yale UniversityIRB required that we pay for any necessary trauma counsel-ing subsequent to the interview. In addition, our researchteam traveled with a trained Rwandan psychologist whoseservices were used only once, for a woman who became up-set after speaking about her experience during the genocidein her individual interview. Participants were also informedabout the protection and confidentiality of their data. Eachparticipant discussed the verbal informed consent twice: oncewhen agreeing to participate in the study, and the secondtime at the end of the year prior to participation in the out-come measurement. The IRB did not require us to debriefparticipants after the unobtrusive behavioral measurementprocedure. Widely adopted ethical standards for IRBs statethat recording behaviors anonymously (without recordingthe names of people enacting the behavior) does not requireinformed consent or debriefing.

When this report was drafted, the first author was atHarvard, where we reapplied to the Harvard IRB for permis-sion to use the data collected with the prison population. Theprison data technically belonged to the NGO. We collectedthese data at the request of the NGO, at the same time andusing the same protocol described previously. However, wehad not obtained permission to work with this populationas academics through the Yale IRB because the board wasconcerned that the prison authorities had reserved the rightto exclude some prisoners from participation. (Their judg-ment ran contrary to our local Rwandan review, where alocal prison advocacy group and several Rwandan academicsreviewed and approved the protocol.) As requested by theNGO, we proceeded with the data collection in the two pris-ons; however, as stipulated by US IRB procedures, we didso as consultants and did not use prison data in academicpresentations or writing. In the prisons, we worked closelywith prison authorities to determine whether any prisonerwas forced to comply with our request to join the study. Itturned out that our study was of no particular concern toprison authorities, who did not attend the radio listening gath-erings and limited their communication with us after ensuringthat high-security prisoners were ineligible for participationin our study. Given that the project was locally endorsedand we observed no interference from prison authorities,we applied to the Harvard IRB to use the prisoner’s datain the analyses for this article. The Harvard IRB reviewedthe case and granted our request. We believed that the datawere important to include because they demonstrate how themedia affects different populations.

APPENDIX II: INDIVIDUAL QUESTIONNAIRE

Introduction (after informed consent): This interview con-tains many statements and opinions given by other people inRwanda. We want to know whether you agree or disagreewith them, and if you agree or disagree somewhat or a lot(four-point scale is repeated after each statement where ap-propriate).

1. I am helpless to prevent violence in my community. (Pro-gram content question)

2. If my colleagues endorse a bad ideology, I cannot changetheir mind by speaking out against it. (Program contentquestion)∗∗Empathy scale: I’m going to ask you if you ever try

to imagine the thoughts or feelings of other peoplewho you don’t know in Rwanda.

3. Do you ever try to imagine the thoughts or feelings ofother people who are prisoners in Rwanda? (When ask-ing this question in the prisons, researchers explained thatthey should respond as though they were speaking aboutprisoners in other prisons, not themselves or their fellowprisoners.)

4. Do you ever try to imagine the thoughts or feelings ofother people who are survivors in Rwanda? (When ask-ing this question in the survivor communities, researchersexplained that they should respond as though they werespeaking about survivors in other communities, not them-selves or their neighbors.)

5. Do you ever try to imagine the thoughts or feelings ofother people who are poor in Rwanda? (Control ques-tion)

6. Do you ever try to imagine the thoughts or feelings ofother people who are leaders in Rwanda?

7. When people marry each other from different (regions,religions, ethnicities, etc.) this contributes to the peace.

8. People who cannot meet their physical and psycholog-ical needs are more likely to blame someone else fortheir problems.

9. One group of people (this is referring to regions, reli-gions, ethnicity, etc.) is often blamed for troubles expe-rienced by the whole country.[Note: This question was not used for analysis because itwas mistranslated in the questionnaire. It was intendedto read as “one group of people is often responsible fortroubles experienced by the whole country.]

10. What if a member of your family dies suddenly, and youhear that that person had a meal at the neighbor’s houseright before he or she died and they poisoned his or herfood. (Program content question)

• What would you think when you heard this story?• What would you do after you heard the story?

11. Violence like the violence that happened in Rwanda in1994 comes about suddenly.

12. Mass violence comes out of small actions, like spreadingnegative ideas about a group of people, or stealing fromthem.

13. Think back to the years of 1993, 1994, and 1995. Did youknow anyone during that time who did or said thingsthat you admire?IF YES: why did you think so?[Note: This question was not used for analysis becauseresearchers experienced difficulties in posing the ques-tion to participants.]

14. If I stand by while others commit evil actions, I am alsoresponsible. (Program content question)

639


15. If I disagree with something that someone is doing orsaying, I should keep quiet.

16. In Rwanda’s history, there has been much violence, butmany people from each group (again, I’m referring toregions, religions, and ethnicity) have not participated inthat violence.[Note: This question was not used for analysis becauseresearchers experienced difficulties in posing the ques-tion to participants.]

17. ∗∗Social distance scale: Think about a person from a dif-ferent religion, region, or ethnicity who has done harmto a person from your group in the past. Would you bewilling toa. Greet this person on the street?b. Work with this person?c. Share a beer (question for males) share food (ques-

tion for females) with this person?d. Allow your child to marry one of his or her children?

18. I advise my children (or the ones I will have in the future)to only marry people from the same group (regions,religions, ethnicities) as theirs.

19. Traumatized people are “mad” (crazy).20. Traumatized people can recover.21. Bad memories and grief can make people seek violent

revenge.22. It is better for my mental health to never talk about the

experiences that have caused me great pain and suffer-ing.

23. I have told someone, like a friend or a family member,about my experiences that have caused me great painand suffering. (yes/no) (Program content question)

• IF NO: I want to tell someone about my memoriesthat caused me great pain and suffering.

24. Perpetrators of violence can also be traumatized by theirown actions.

25. Someone has talked to me about their experiences thathave caused them great pain and suffering.

26. Recovering from grief (intimba) and from trauma(ihungabana) may take a very long time.

27. There is mistrust (ukutizerana) in my village.28. It is naıve to trust people.29. A pregnant woman who has AIDS can be given a chance

to have a healthy baby.30. You can safely share something with someone who has

AIDS.31. It is necessary that every woman who is pregnant go to

the health center to be tested.

APPENDIX III: UNFINISHED RADIO SCENEUSED FOR ROLE PLAY

NARRATOR: We are in the village of MAHINYUZAat the border. Some of the villagers of Mahinyuza havestarted to discuss the rumor they have heard that the vil-lagers of MAKOBA in the neighboring country are goingto migrate from their village, fleeing a rampant famine thatis going to kill them. KANDONYI, MUKAMA, ANDCACANA are discussing this news, but there are otherpeople around. [For women’s version: After the conver-sation between the three men is over, I would like to askthe women present here to give comments on the scene,and after, to take a role and finish the story.]

KANDONYI: Mukama, do you know why I have come to seeyou? Although I know that this rampant famine didn’t spare

you either, please do your best and find me one thousandfrancs to lend me, to buy beans for my children. They didn’teat last night!

MUKAMA: Even me, I’m so poor and hungry that when Isneeze, my teeth fall out. But let me give you the last twohundred that I have. (Starts to check in his pockets) My fatherRwabuzisoni gave me a cow! Where is the money that I hadfrom that?

KANDONYI: Do you mean you can’t find it? Please checkwell so that can I go and save my children before hunger killsthem.

MUKAMA: Where is it? When I left Makoba I had it!Maybe it’s the people of Makoba who stole it! I don’t trustthem.

CACANA: (Runs up to them, out of breath) You, people atMukama’s place!

MUKAMA: (Surprised) Cacana, what happened to makeyou run so fast you cut through trees and stones?

KANDONYI: Cacana, is the situation peaceful?

CACANA: I have passed here to tell you what I have heardin the pub so that we can consider what we can do.

MUKAMA: (Worriedly) What’s that you have heard in thepub?

KANDONYI: Let it be of sweat and not of blood!

CACANA: (Fearfully) I have heard that our neighbors ofMakoba have packed all their things to come settle here,because in their village it takes five men to close a door (aKinyarwanda proverb meaning they’re weak and hungry).

MUKAMA: (Shocked) My goodness! They want to migratehere! Do they think that we were spared from the famine!Even us, we’re so hungry that when the flies buzz around oureyes we try to kill them and eat them. They have to look forsomewhere else instead of coming here to finish the little wehave!

CACANA: I have come to tell you the measures we havealready taken with the people who have told me that. Wehave decided to go to the border and stop them from comingto Mahinyuza. Because you know the problem we have here.

KANDONYI: Yah, babababababa. (exclamation) Why doyou think like that? Instead of looking at how to receive andwelcome them warmly as neighbors, you are setting up tricksto stop them!

MUKAMA: Can you believe what this man is saying! Whilewe are crying, you are singing! If we let them come, what willyou eat? You were saying that your children are sufferingfrom hunger, and you are borrowing money! They can’t puttheir feet on this soil, they have to go elsewhere!

CACANA: You are delaying! And if we delay, they will behere in a very short time.

KANDONYI: Let’s welcome them warmly and share thelittle we have. They may save us in the future, you neverknow. What if you stop them and they fight you?

CACANA: Fighting us! Are they stronger than we? Mukamaand I are resolved to arm ourselves with our traditionalweapons, so that if they try to force entry into our village,we’ll make them tremble with our threats.

KANDONYI: What you want to do is not good. That’s vio-lence.

MUKAMA: (Talks as he leaves) Caca, let me get my swordfrom its girdle and chase them away. (His voice becomes more

640


distant as he leaves.) Maybe it’s they who took my money,as I was saying! It’s possible it’s them! They were all madefrom the same hammer. Leave alone Kandonyi who is talkingnonsense!

APPENDIX IV: CALCULATION OFROLE-PLAY STATISTICS

We describe two methods for calculating the relative fre-quency of each code for reconciliation and health role playsin comparison to the total number of codes. First, let Ck

denote the code assigned to line k. We are interested in therelative frequency with which lines are assigned to code k = Kand, in particular, the way this rate varies between treatmentand control groups. One simple approach is to calculate thetotal number of k = K codes in each experimental groupand divide by the sum of all codes assigned to those lines.For example, the rate of k = K codes in the reconciliation

group across all sites s, all groups g within those sites, and allversions v enacted by each group is

Tunweighted%codea =∑

CaGV∑

CSGV

An alternative approach calculates the percentage of codesk = K within each site and then averages these percentagesacross sites. Thus, for example, the rate in the reconciliationgroup would be calculated as

S%codea =∑

CaGV∑

CGV,

where the number of sites S is 7. The main difference betweenthese two approaches is that the latter weights all sites equally,whereas the former weights the data so that all line codeshave equal weight.

APPENDIX V: SUPPLEMENTAL TABLES

TABLE A1. Ratio of Community-Based and Government/NGO-BasedSolutions Proposed per Scene in Reconciliation and Health Role Plays

Origin of Assistance Given to Refugees

No. of Scenes Community (Ratio) Gov/NGO (Ratio)

ReconciliationNorthwest 5 4 .8 0 0South 6 6 1 0 0Prison 3 1 .33 0 0Survivor 4 5 1.25 0 0HealthNorthwest 4 1 .25 0 0South 5 2 .4 6 1.2Prison 4 1 .25 4 1Survivor 2 0 0 4 2

NGO, nongovernmental organization.Note: Entries are counts or ratios of actions from the role plays conducted in each research site.

641


TABLE A2. Conditional Effects of RadioTreatment on Dissent: Interactions with Age,Experience of Violence, and Access to Media

Dissent

Reconciliation −.035 .242 −.217program (.286) (.394) (.601)

Age −.007(.004)

Age × .008reconciliation (.007)

Experience of .068violence (.182)

Experience of violence × .028reconciliation (.220)

Access to media −.206(.183)

Access to media × .272reconciliation (.313)

Northwest −.207 −.246∗ −.276∗

(.108) (.119) (.123)Northeast −.241∗ −.252∗∗ −.286∗∗

(.078) (.031) (.020)South −.192∗ −.206∗∗ −.216∗∗

(.087) (.051) (.063)Twa −.329∗ −.292∗ −.331∗

(.128) (.117) (.124)Prison −.131 −.159 −.139

(.151) (.148) (.287)Survivor .055 −.055 −.050

(.069) (.062) (.036)Cut 1 −.919 −.552 −1.066

(.192) (.293) (.360)Cut 2 −.469 −.116 −.631

(.174) (.297) (.356)Cut 3 −.142 .221 −.297

(.165) (.303) (.368)Log likelihood −616.15 −553.08 −555.54

Notes: ∗p < .05, ∗∗p < .001, two tailed. None of the estimatedinteraction effects approaches statistical significance. Dissentis measured using a survey item that ranges from 1 (I shouldstay quiet) to 4 (I should dissent). Reconciliation is scored 1 iffor the reconciliation program and 0 for the health program. Ageis measured in years. Experience of violence is coded 1 = nofamily members lost and 2 = yes, family members lost. Accessto media is coded 1 = does not own a radio and 2 = owns radio.

REFERENCESAdorno, Theodor W., Else Frenkel-Brunswik, Daniel J. Levinson,

and R. Nevitt Sanford. 1950. The Authoritarian Personality. NewYork: Harper.

African Rights. 1995. Rwanda: Death, Despair and Defiance.London: African Rights.

Almond, Gabriel, and Sidney Verba. 1963. The Civic Culture: Politi-cal Attitudes and Democracy in Five Nations. Newbury Park, CA:Sage.

Andre, Catherine, and Jean-Phillipe Platteau. 1998. “Land Rela-tions under Unbearable Stress: Rwanda Caught in the MalthusianTrap.” Journal of Economic Behavior and Organization 34 (Jan-uary): 1–47.

Arceneaux, Kevin. 2005. “Using Cluster Randomized Field Ex-periments to Study Voting Behavior.” Annals of the AmericanAcademy of Political and Social Science 601 (September): 169–79.

TABLE A3. Conditional Effects of RadioTreatment on Social Distance According toAge, Experience of Violence, and Access toMedia

Social Distance

Reconciliation .366 −.174 −.590program (.199) (.220) (.492)

Age .001(.005)

Age × −.010reconciliation (.006)

Experience of .014violence (.146)

Experience of violence × .061reconciliation (.134)

Access to media .179(.177)

Access to media × .281reconciliation (.261)

Northwest .046 .064 .085(.060) (.112) (.103)

Northeast .288∗ .311∗ .297∗

(.087) (.136) (.112)South .209∗ .188 .240∗

(.067) (.123) (.102)Twa −.078 −.038 .009

(.272) (.279) (.266)Prison .247∗ .258 .282

(.110) (.154) (.140)Survivor −.279∗ −.254 −.250∗

(.080) (.131) (.099)Constant 3.258 3.255 2.929

(.202) (.193) (.384)

Notes: ∗p < .05, ∗∗p < .001, two tailed. None of the estimatedinteraction effects approaches statistical significance. SeeTable 4 for scoring of social distance scale, which is the sumof four survey responses. Reconciliation is scored 1 if for thereconciliation program and 0 for the health program. Age ismeasured in years. Experience of violence is coded 1 = nofamily members lost and 2 = yes, family members lost. Accessto media is coded 1 = does not own a radio and 2 = owns radio.

Arendt, Hannah. 1963. Eichmann in Jerusalem: A Report on theBanality of Evil. New York: Penguin Books.

Asch, Solomon. 1956. Social Psychology. Englewood Cliffs, NJ:Prentice-Hall.

Bandura, Albert. 2000. “Exercise of Human Agency through Col-lective Efficacy.” Current Directions in Psychological Science 9(June): 75–78.

Bandura, Albert. 2002. “Environmental Sustainability by So-ciocognitive Deceleration of Population Growth. In The Psy-chology of Sustainable Development, eds. Peter Schmuch andWesley Schultz. Dordrecht, The Netherlands: Kluwer; 209–38.

Blattman, Christopher. 2008. “From Violence to Voting: War andPolitical Participation in Uganda.” Working Paper 138, Center forGlobal Development.

Boal, Augusto. 1992. Games for Actors and Non-actors, Trans. AdrianJackson. London: Routledge.

642


Bond, Rod, and Peter B. Smith. 1996. “Culture and Conformity: AMeta-Analysis of Studies Using Asch’s 1952b, 1956 Line JudgmentTask.” Psychological Bulletin 119 (January): 111–37.

Breed, Ananda. 2007. “Performing Gacaca in Rwanda: Local Cul-ture for Justice and Reconciliation.” In The Cultures and Global-ization Series—Vol. 1: Conflicts and Tensions, eds. Helmut Anheierand Yudhishthir R. Isar. London: Sage; 306–12.

Browning, Christopher. 1993. Ordinary Men: Reserve Police Battal-ion 101 and the Final Solution in Poland. New York: Harper.

Chalk, Frank, and Kurt Jonassohn. 1990. The History and Sociologyof Genocide. New Haven, CT: Yale University Press.

Chandler, David. 1999. Voices from S-21: Terror and History in PolPot’s Secret Prison. Berkeley: University of California Press.

Chwe, Michael. 2001. Rational Ritual: Culture, Coordination, andCommon Knowledge. Princeton, NJ: Princeton University Press.

Craig, Gordon A. 1980. The Germans. New York: New AmericanLibrary.

Dallaire, Romeo A., and Brent Beardsley. 2003. Shake Hands withthe Devil: The Failure of Humanity in Rwanda. New York: Carroll& Graf.

Douglas, Mary, and Aaron Wildavsky. 1982. Risk and Culture.Berkeley: University of California Press.

Eckstein, Harry. 1988. “A Culturalist Theory of Political Change.”The American Political Science Review 82 (September): 789–804.

Eckstein, Harry. 1992. Regarding Politics: Essays on Political Theory,Stability, and Change. Berkeley: University of California Press.

Fisher, Suzanne. 2004. “Tuning into Different Wavelengths: Lis-tener Clubs for Effective Rwandan Reconciliation Radio Pro-grammes.” Paper presented to the Fourth International Confer-ence on Entertainment-Education and Social Change, Cape Town,South Africa.

Freedman, David A. 2006. “Statistical Models for Causation: WhatInferential Leverage Do They Provide?” Evaluation Review 30(December): 691–713.

Fuji, Lee Ann. 2009. Killing Neighbors: Social Dimensions of Geno-cide in Rwanda. Ithaca, NY: Cornell University Press.

Gabisirege, S., and Stella Babalola. 2001. Perceptions about theGacaca Law in Rwanda: Evidence from a Multi-Method Study.Special Publication No. 19. Baltimore: Johns Hopkins UniversitySchool of Public Health, Center for Communication Programs.

Galvan, Dennis C. 2004. The State Must Be Our Master of Fire:How Peasants Craft Culturally Sustainable Development in Sene-gal. Berkeley: University of California Press.

Gerbner, George, Larry Gross, Michael Morgan, Nancy Signorielli,and James Shanahan. 2002. “Growing Up with Television: Cul-tivation Processes.” In Perspectives on Media Effects: Advancesin Theory and Research. 2nd ed., eds. Jennings Bryant and DolfZillmann. Mahwah, NJ: Lawrence Erlbaum; 43–67.

Goldhagen, D. J. 1996. Hitler’s Willing Executioners: Ordinary Ger-mans and the Holocaust. New York: Alfred A. Knopf.

Gourevitch, Philip. 1998. We Wish to Inform You Tomorrow We WillBe Killed with Our Families: Stories from Rwanda. New York:Farrar Straus & Giroux.

Gurr, Ted Robert. 1970. Why Men Rebel. Princeton, NJ: PrincetonUniversity Press.

Heath, Robert, and Jennings Bryant. 1992. Human Communica-tion Theory and Research: Concepts, Contexts, and Challenges.Hillsdale, NJ: Lawrence Erlbaum.

Hinton, Alexander L. 2005. Why Did They Kill? Cambodia in theShadow of Genocide. Berkeley: University of California Press.

Huntington, Samuel P. 2000. “Cultures Count.” In Culture Matters,eds. S.P. Huntington and L.E. Harrison. New York: Basic Books;xiii–xvi.

Inglehart, Ronald F. 1997. Modernization and Postmodernization.Princeton, NJ: Princeton University Press.

Iyengar, Shanto. 1991. Is Anyone Responsible? How TelevisionFrames Political Issues. Chicago: University of Chicago Press.

Jackman, Robert W., and Ross A. Miller. 2004. Before Norms. AnnArbor: University of Michigan Press.

Katz, Elihu, and Paul Lazarsfeld. 1955. Personal Influence: The PartPlayed by People in the Flow of Mass Communication. New York:Free Press.

Kelman, Herbert, and Lee Hamilton. 1989. Crimes of Obedience.New Haven, CT: Yale University Press.

Klemperer, Victor. 1998. I Will Bear Witness. New York: RandomHouse.

Lacey, Marc. 2004. “A Decade After Massacres, Rwanda OutlawsEthnicity.” New York Times, 9 April.

Laitin, David. 1986. Hegemony and Culture: The Politics of Reli-gious Change among the Yoruba. Chicago: University of ChicagoPress.

Lemarchand, Rene. 1970. Rwanda and Burundi. London: Pall Mall.Lewin, Kurt. 1948. Resolving Social Conflicts. New York: Harper and

Brothers.Lyons, Robert, and Scott Straus. 2006. Intimate Enemy: Images and

Voices of the Rwandan Genocide. New York: Zone Books.Milgram, Stanley. 1974. Obedience to Authority: An Experimental

View. New York: Harper & Row.Miller, Dale, Benoit Monin, and Deborah Prentice. 2000. “Plural-

istic Ignorance and Inconsistency Between Private Attitudes andPublic Behavior.” In Attitudes, Behavior, and Social Context: TheRole of Norms and Group Membership, eds. Deborah J. Terry andMichael A. Hogg. Mahwah, NJ: Erlbaum; 95–113.

Newbury, Catharine. 1988. The Cohesion of Oppression: Clientshipand Ethnicity in Rwanda, 1860–1960. New York: Columbia Uni-versity Press.

Oxford, English Dictionary. 2nd ed. 20 vols. Oxford: Oxford Univer-sity Press, 1989.

Packer, Dominic J. 2008. “Identifying Systematic Disobedience inMilgram’s Obedience Experiments: A Meta-Analytic Review.”Perspectives on Psychological Science 3 (July): 301–4.

Paluck, Elizabeth L. 2006. The Second Year of a “New Dawn”: YearTwo Evidence for the Impact of the Rwandan Reconciliation RadioDrama Musekeweya. Kigali: LaBenevolencija.

Paluck, Elizabeth L. 2009.“Reducing Intergroup Prejudice and Con-flict Using the Media: A Field Experiment in Rwanda.” Journal ofPersonality and Social Psychology 96 (March): 574–87.

Parsons, Talcott. 1951. The Social System. New York: Free Press.Prendergast, John, and David Smock. 1999. Post-Genocidal Recon-

struction: Building Peace in Rwanda and Burundi. Special ReportNo. 53. United States Institute of Peace USI.

Prewitt, Kenneth. 1975. “Some Doubts about Political SocializationResearch.” Comparative Education Review 19 (February): 105–14.

Prunier, Gerard. 1995. The Rwanda Crisis: History of a Genocide.New York: Columbia University Press.

Putnam, Robert. 1993. Making Democracy Work: Civic Traditions inModern Italy. Princeton, NJ: Princeton University Press.

Pye, Lucian W., and Sidney Verba, eds. 1965. Political Culture andPolitical Development. Princeton, NJ: Princeton University Press.

Ramsauer, Ernest, Jr. 1957. The Young Turks: Prelude to the Revo-lution of 1908. New York: Russell & Russell.

Republic of Rwanda. 2001. National Decentralization Policy [May2001 version]. Kigali: Ministry of Local Government and SocialAffairs.

Republic of Rwanda. 2002. Poverty Reduction Strategy Paper. Kigali:Ministry of Finance and Economic Planning, National PovertyReduction Programme.

Robinson, Paul H., and John M. Darley. 2003. “The Role of Deter-rence in the Formulation of Criminal Law Rules.” GeorgetownLaw Journal 91: 949–1002.

Rogowski, R. 1974. Rational Legitimacy: A Theory of Political Sup-port. Princeton, NJ: Princeton University Press.

Rohrschneider, Robert. 1999. Learning Democracy: Democratic andEconomic Values in Unified Germany. Oxford: Oxford UniversityPress.

Rosin, H. 2006. “Life Lessons.” New Yorker Magazine 82 (June 5):40–45.

Schatzberg, Michael. 2001. Political Legitimacy in Middle Africa:Father, Family, Food. Bloomington and Indianapolis Indiana Uni-versity Press.

Scott, James. 1990. Domination and the Arts of Resistance: HiddenTranscripts. New Haven, CT: Yale University Press.

Sears, David O., and Carolyn L. Funk. 1991. “The Role of Self-Interest in Social and Political Attitudes.” Advances in Experi-mental Social Psychology 24: 1–91.

Sewell, William H., Jr. 2005. “The Concept(s) of Culture.” In Log-ics of History, ed. William H. Sewell, Jr. Chicago: University ofChicago Press; 152–74.

643


Shils, Edward A. 1972. The Constitution of Society. Chicago: Univer-sity of Chicago Press.

Singhal, A., M. J. Cody, E. M. Rogers, and M. Sabido, eds. 2004.Entertainment-Education and Social Change: History, Research,and Practice. Mahwah, NJ: Lawrence Erlbaum.

Singhal, Arvind, and Everett M. Rogers. 2004. “The Statusof Entertainment Education Worldwide.” In Entertainment-Education and Social Change: History, Research, and Prac-tice, eds. Arvind Singhal, Michael J. Cody, Everett M. Rogers,and Miguel Sabido. Mahwah, NJ: Lawrence Erlbaum; 3–18.

Straus, Scott. 2006. The Order of Genocide: Race, Power, and War inRwanda. Ithaca, NY: Cornell University Press.

Straus, Scott. 2007. “What Is the Relationship between Hate Radioand Violence? Rethinking Rwanda’s ‘Radio Machete.’” Politicsand Society 35 (4): 609–37.

Swidler, Ann. 1986. “Culture in Action: Symbols and Strate-gies.” American Sociological Review 51 (April): 273–86.

Transparency International. 2008. “Transitional Justice: No Impunityfor Corruption and Other Crimes of Greed.” September 12.www.transparency.org/news_room/latest_news/press_releases_nc/2008/2008_09_12_ti_zimbabwe_trans_just. (July 30, 2009).

United Nations. 1999. Resolutions Adopted by the General Assembly:53/243. Declaration and Programme of Action on a Culture ofPeace. New York: United Nations.

Verwimp, Philip. 2005. “An Economic Profile of Peasant Perpetra-tors of Genocide: Micro-Level Evidence from Rwanda.” Journalof Development Economics 77 (August): 297–323.

Wantchekon, Leonard. 2003. “Clientalism and Voting Behavior:Evidence from a Field Experiment in Benin.” World Politics 55(April): 399–422.

Weber, Max. 1978. Economy and Society. Berkeley: University ofCalifornia Press.

Wedeen, Lisa. 1999. Ambiguities of Domination: Politics, Rhetoric,and Symbols in Contemporary Syria. Chicago: University ofChicago Press.

Wedeen, Lisa. 2002. “Conceptualizing Culture: Possibilities for Polit-ical Science.” American Political Science Review 96 (December):713–28.

Wilson, Richard W. 2000. “The Many Voices of Political Culture:Assessing Different Approaches.” World Politics 52 (January):246–73.

Zorbas, Eugenia. 2007. Reconciliation in Post-Genocide Rwanda:Discourse and Practice. Unpublished PhD thesis. London Schoolof Economics.

644

Documents

Deference, Dissent, and Dispute Resolution: An ... · Deference, Dissent, and Dispute Resolution November 2009 The article is structured as follows. We begin by discussing media interventions