Reducing intergroup prejudice and conflict with the mass media A

Document Sample
Reducing intergroup prejudice and conflict with the mass media A Powered By Docstoc
					Reducing intergroup prejudice and conflict with the mass media: A field experiment in

                                    A Dissertation
                   Presented to the Faculty of the Graduate School
                                   Yale University
                           in Candidacy for the Degree of
                                Doctor of Philosophy

                               Elizabeth Levy Paluck

                       Dissertation Director: Donald P. Green

                                     May 2007


 Reducing intergroup prejudice and conflict with the mass media: A field experiment in

                                  Elizabeth Levy Paluck


Can the media reduce intergroup prejudice and conflict? Despite the high stakes of this

question, our understanding of the mass media’s role in shaping attitudes and behaviors,

especially prejudiced attitudes and behaviors, is very limited. This study attempted to fill

this gap with the first experimental evaluation of a radio program’s impact on intergroup

prejudice and conflict in a real world setting. The study took place in the East African

country of Rwanda where a war and genocide resulted in the deaths of more than 10% of

the population over the course of 3 months in 1994. I randomly assigned Rwandan

communities to listen to one of two radio programs: a reconciliation radio soap opera and

a health soap opera. The reconciliation soap opera aimed to influence beliefs about

intergroup prejudice, mass violence and trauma with a series of educational messages,

and its fictional characters portrayed positive social norms regarding intergroup behavior

and trauma healing. Study participants listened to the programs over the course of one

year, during which I collected observational data on their discussions of and emotional

reactions to the programs. At the end of the year I measured outcomes with standardized

questionnaires, focus groups, and behavioral observations. The reconciliation radio

program did little to influence listeners’ beliefs about the radio program’s educational

messages. However, results support the hypothesis that radio programs can influence

behaviors and perceived social norms regarding intergroup relations and trauma healing.

The reconciliation program affected listeners’ perceptions of and behaviors toward some

of the most critical issues for Rwanda’s post conflict society, such as intermarriage, open

dissent, trust, and talking about personal trauma. A pattern of perceived norm and

behavior change was observed across measures of participants’ attitudes, group

discussions, and behaviors during deliberations about a communal resource. The program

also increased empathy for other Rwandans. Taken together, the results suggest that radio

can communicate social norms and influence behaviors that contribute to intergroup

tolerance and reconciliation. Field experiments like this one can be deployed to measure

the causal impact of the mass media, and to animate theoretical work on the processes of

media influence and prejudice and conflict reduction.

 2007 by Elizabeth Levy Paluck
      All rights reserved.

                                   Table of contents




Reducing Prejudice and Intergroup Conflict with Media: Interventions, Theories, and


The Present Research........…….……….………………………………………………...29

Study Method....................…….……….………………………………………………...43

Data Collection.................…….……….…………………………………………….......53





Appendix A......................…….……….………………………………………………..119

Appendix B......................…….……….……………....………………………………..121

Appendix C......................…….……….………………………………………………..124


                                                Figures and Tables

Figure 1: Map of research sites……….……………….…………………………………44

Table 1: Sample pretest measures demonstrating balance between pairs of


Figure 2: Pictorial agreement scale........……………………….………………………...55

Table 2: Beliefs about violence are unaffected.....….……………………………………63

Table 3: Focus group opinions about factors that contribute to violent ideologies

       in a society..............................…….…………….….....…………………………65

Table 4: Beliefs about intermarriage bringing peace are unaffected.……………………67

Table 5: Beliefs about trauma symptoms are not affected in the

       predicted direction.............................................................................................…68

Table 6: Prescriptive norm about intermarriage is affected..…………………………....70

Table 7: Norms about trust and dissent in the community are affected....……………....73

Table 8: Norm of talking about trauma is affected.....……………………………..........74

Table 9: Beliefs are unaffected and norms are affected by the health

       soap opera..............................................................................................................75

Table 10: Desire for social distance from “those who have harmed you or

       your group” is unaffected......................................................................................76

Table 11: Increased empathy for real life Rwandans who were portrayed

       in the radio soap.................................................................................................…79

For Anne and Alan Paluck, who gave me life more than once


        I researched and wrote this dissertation with the trust, generosity, experience and
wisdom of people on three different continents. Their individual and collective
contributions shaped the dissertation as well as the course of my life—the three-year
duration of the project and the future that awaits me now. Acknowledging them on this
page is just one way I want to express my deep gratitude.

        In Europe, one of my first supporters was George Weiss, the founder and creative
inspiration of LaBenevolencija. George responded to my initial inquiries about studying
the radio program and after a few phone calls decided to take his chances on a sure-
sounding but relatively inexperienced graduate student. I am grateful for his confidence,
energy, and friendship. LaBenevolencija also devoted substantial financial support to the
project, which Robin Brinster coordinated from Amsterdam along with Anneke van

        In the US, while I was conceptualizing the project Paul Bolton, Charles Mironko,
Scott Straus, and Harvey Weinstein gave me important advice from their own deep
experiences with Rwanda and with post-conflict research. Their generosity set the
standard for my future participation in that community of scholars. Without grants from
the John Perry Miller Fund, the Yale Globalization Center, and the Institute for Social
and Policy Studies (ISPS) at Yale University, and continuous support from a National
Science Foundation Graduate Fellowship, my research would never have taken flight.
And when bank channels and SWIFT codes between the US and Rwanda failed, Pamela
LaMonaca and Barbara Dozier at ISPS worked tirelessly to throw me a line. Their office
was not only a place where all my administrative problems were handled with grace and
cheer, it was also a place for good conversation and chocolate.

       I was privileged to use the Yale libraries; in particular, Julie Linden at the Yale
Social Science Library was invaluable for help with complicated fact searches. Kyle
Hood in the Statistics Laboratory at Yale (StatLab) solved the thornier problems of my
data management and analysis, and Kevin Arceneaux generously lent me his expertise
with respect to cluster analysis. Themba Flowers gave me an indefinite leave from my
own StatLab shifts and joined Nathan Paluck, Elizabeth Vozzola, Amelia Shillingford,
and many friends and family in cheering me on from afar.

        When the time came to read drafts, my dissertation committee did so thoroughly
and critically. I am grateful that John Bargh, Peggy Clark, Richard Eibach and Marianne
LaFrance pushed me to explore in greater theoretical detail the social psychological
underpinnings of radio listening. Peter Salovey, the co-chair of the committee, took time
out from being Dean of Yale College to engage my project with his remarkable wisdom,
energy and humor. At various points, Lee Ann Fujii, Izzat Jarudi, Shivaji Mukherjee,
Stathis Kalyvas, Nicholas Sambanis, Scott Straus, Eric Uhlmann, and Philip Verwimp
read drafts and offered comments that both improved my thinking and reassured me that I
was on the right track, if such smart people approved of my general ideas. I presented this

work and received feedback at the Comparative Politics Workshop at Yale University,
the four-school Psychology Graduate Conference at New York University, the UCLA
Psychology Department, and at the BA Festival of Science in Dublin. I’m glad that David
Sears shared his thoughts with me about how to frame my findings before I finished the
final draft. Michael Kavanagh edited and commented on versions from the first to the
last. His influence on my writing, my thinking, and my quality of life has been one of the
greatest gifts of my graduate years. I’m glad we met for tea that rainy afternoon in Kigali.

       In Rwanda, an extended circle of Rwandan and expatriate friends and colleagues
made the country feel as familiar and as beloved as a second home. Without these people
my experience of Rwanda would have been much more superficial, not to mention
boring: Solange Ayanone, Lie Ayanone, Heather Baldwin, Nadège Degris, Klaas
DeJonge, Lee Ann Fujii, Gabriel Gabiro (and family, particularly Zula and Mama Gabi),
Riccardo Gangale, Veronique Geoffrey, Beza Gisèle, Willy Kamanzi, Thomas Kamilindi,
Jean Karambizi, Jean Charles Paras, Laurette Rudasingwa, Virgile Uzabumagabo, Philip
Verwimp, Ante Vrijlandt, Lars Waldorf, Julie Weaver, Sarah Wells, and Eugenia Zorbas.
Each person deserves a line of his or her own.

        I must acknowledge the cooperation of CAURWA, IBUKA, and the Rwandan
Ministries, particularly MINALOC. Mr. Christophe Bazivamo and Mr. Charles
Murigande provided critical support for the entire radio project, without which there
would have been no research. I also received permissions to conduct the field research
from every relevant provincial prefect, mayor, responsable and other Rwandan
authorities. I am grateful for every stamped paper I accumulated that made this work

         My colleagues at LaBenevolencija were inspirational and deeply enriching to
work with. Suzanne Fisher, the first Chef de Mission of LaBenevolencija, handed me her
list of contacts and familiarized me with entertainment-education radio production. Ervin
Staub and Laurie Pearlman agreed with George Weiss that I could study the radio project,
to which their theories were central. The LaBenevolencija writers created a rich and
popular program to work with: Perpetue Mukahigiro and the head writer Charles
Rukundo, who also assisted with writing and production for parts of the field evaluation.
Jean Karambizi was a consultant par excellence for LaBenevolencija, and taught me a
great deal about the Great Lakes and how to milk a cow. Bernard Bariyanga led me
through the intricacies of Rwandan banks, taxes, and contracts. I learned and laughed a
great deal working with Johan Deflander, the second Chef de Mission of
LaBenevolencija, who should write a book about his experiences working with radio
throughout Africa.

         Marie Lequin Coutin, Narcisse Kalisa, Josephine Uwamaryia and the Urunana
staff from BBC/Health Unlimited generously allowed me to use Urunana as our
comparison radio program for the evaluation. Theoneste Rutagengwa, my first research
assistant, patiently and intelligently helped me to set up the structure of the evaluation.
Justin Rwibasira, Antoinette Umugwaneza, and Theotime Taremana were the faithful
research assistants who brought Musekeweya to participants in the field and filed

observational notes each week. The translator I worked with during my first trip to
Rwanda was Leonard Lengema, who traveled with me to every single province besides
Umutara (in the old provincial system) and endured endless questions. The indefatigable
translators for some parts of my interview data were Sylvestre Ndahayo and Olivier
Mpumuro. On his own time, Olivier taught me a great deal about Kinyarwanda,
sensitively parsing differences in denotation and connotation for various words used in
our interviews.

        The research team became my family out in the field: Elévanie Bayisenge,
Geneviève Mukandekezi, Odette Mukeshimana, Evariste Nkunda, Marie Louise
Nyinawumuntu, Rosalyn Nzamramba, Louise Ruberwa, Jacques Sindayigaya, Teotime
Taremara, Anicet Twagirumukiza, Diane Uwera, and especially the “rapid response
team” that I relied upon until the very end: Willy Kamanzi, Alice Kambayire, Ildephonse
Nizeyimana, Fidele Rubayiza, Laurette Rudasingwa, Eric Ruhamirira, Prosper Ruvusha,
Justin Rwibasira, and Virgile Uzabumagabo. Virgile sat next to me during translation of
the interviews and focus groups and made sure that nothing was lost and everything was
placed into a rich context. My most interesting insights into Rwandan culture, especially
its proverbs and stories, come from time spent with him.

        I owe a great debt to the more than 500 ordinary Rwandans who participated in
this research project. These individuals’ willingness to share parts of their current lives
and past memories involved a measure of sacrifice and bravery that I will never forget
and cannot repay in full. I am filled with respect and gratitude for their fidelity to the
project and for the insights they contributed.

         Albert Nzamukwereka was the research assistant who saw this project through
from the pretest to the final round of data entry. Albert traveled the country with me,
pursued research permissions with countless meetings and phone calls, introduced me to
dozens of people, and while I was away handled everything from finances to personnel
supervision to mechanical breakdowns. Albert was adept at so many things, but above all
his gift for interacting with people assured the success of the project. He is a friend and a
colleague for life and a great deal of the credit for the execution of this project belongs to

        Donald P. Green was my dissertation director and academic advisor. Don was the
person who challenged me to find out exactly what psychologists know about reducing
prejudice and conflict in the real world. From that moment on, his unwavering support
allowed me the freedom to be daring and single minded in my search for the answers. It
is because of Don’s intelligence, generosity, his leadership in the area of field
experimentation, commitment to interdisciplinary approaches, and keen attention to
strategy and detail that my dissertation was possible. Don gave me a great deal of
independence, but when asked he would attend to anything from my grammar to my
overarching career path. My regard for him has evolved from a timid admiration to a
deep collegial appreciation. I have been extraordinarily fortunate to be his student.


       For nearly a century, psychological research has tackled the societal problems of

prejudice and intergroup conflict. Few topics have attracted a greater range of theoretical

perspectives. Theories implicate personality traits (Adorno, Frenkel-Brunswik, Levinson,

Sanford, & Nevitt, 1950), displaced aggression (Hovland & Sears, 1940), scarcity of

material resources and status (Sherif & Sherif, 1953; Blumer, 1958), homophily and

ingroup favoritism (Tajfel & Turner, 1979), various combinations of conservative or

liberal values and negative affect (Gaertner & Dovidio, 1986; Katz & Hass, 1988; Kinder

& Sears, 1981), human proclivities for hierarchy (Sidanius & Pratto, 1999), and

automatic habits of the mind (Banaji, 2001a; Devine, 1989). Although this literature has

generated some promising theoretical and empirical leads, scholars do not currently have

a clear answer to the question “What interventions have been shown to reduce prejudice

and conflict in real world settings?”

       Our understanding of the mass media’s role in shaping attitudes and behaviors,

especially prejudiced attitudes and behaviors, is even more limited. Spurred by

widespread use of propaganda in the first and second world wars (Hovland, Lumsdaine,

& Sheffield, 1949; Lippman, 1922; Lewin, 1952), early psychological research explored

how the media could play a role in fomenting (Doob, 1935; Lasswell, 1927) and reducing

prejudice and conflict (Cooper & Jahoda, 1947; Flowerman, 1949; Peterson & Thurstone,

1933). This research began the work of identifying important theoretical issues, but soon

after psychologists drifted away from the study of the media.

       The research presented here is an attempt to fill gaps in both literatures. The

setting for this study is the East African country of Rwanda where a war and genocide

resulted in the deaths of more than 10% of the population and 75% of the Tutsi ethnic

minority population over the course of 3 months in 1994. This study reports the results of

a randomized field experiment conducted with the non-governmental organization

LaBenevolencija, which ten years after the genocide produced a yearlong radio soap

opera designed to promote reconciliation in Rwanda.

       I assessed the radio program’s impact on listeners’ beliefs (regarding the

program’s educational messages about prejudice, violence and trauma) and on listeners’

perceptions of social norms (relevant to those depicted by fictional characters in the

program). In other words, this study tested two related propositions: the mass media have

the capacity to convey a particular set of beliefs (in this case, beliefs about prejudice

reduction, violence prevention and trauma healing), and to promote new social norms that

describe and prescribe peaceful intergroup relations and reconciliation.

       A fundamental goal of this research was to collect rigorous real world data that

can speak to the causal impact of the media. Although the mass media have been a

longstanding object of social scientific interest, the number of randomized field

experiments that have assessed the causal impact of television, radio, or newspapers is

very small. Nearly every study in this field relies on observational data or lab

experimentation; rarely are the effects of a media intervention gauged by randomly

assigning media exposure outside a laboratory setting. I randomly assigned Rwandan

communities to listen to one of two radio programs: the reconciliation radio program or a

similarly designed radio program dealing with health issues. After one year of exposure,

I measured outcomes with a combination of standardized questionnaires, focus groups,

and planned activities that allowed for behavioral observation.

       An interesting pattern of mixed effects emerged from the study. The

reconciliation radio program did little to influence beliefs about intergroup relations and

trauma. Reconciliation listeners’ endorsement of the descriptive theoretical principles that

guided the program—formalized into educational messages about how prejudice,

violence and trauma develop and can be reduced—was no different from that of health

program listeners. However, the study’s results support the hypothesis that radio

programs can communicate and reinforce positive social norms and behaviors regarding

intergroup relations and trauma healing. The radio program affected listeners’

perceptions of and behaviors toward some of the most critical issues for Rwanda’s post

conflict society, such as ethnic intermarriage, open dissent, empathy and trust. This

pattern of effects was manifested in self-reported attitudes, group discussions, and

behavior in collective-decision making. On the whole the results strongly suggest that

radio can effectively communicate social norms that contribute to intergroup tolerance

and reconciliation.

       The radio intervention was multifaceted, which raises questions regarding which

of the many factors at play were responsible for the changes in participants’ perceived

norms and behaviors. I discuss such factors highlighted in the current study, such as

group discussion, humor and imagination, which are part of the social and

phenomenological experience of media consumption in the real world. The current

investigation offers many ideas for future research for exploring the social psychological

mechanisms of prejudice and conflict reduction with the mass media.

       The dissertation is structured as follows. First I describe the range of media

interventions that exist to reduce prejudice and conflict, the theories psychologists

mobilize to understand them, and the accumulated evidence for media effects. We see

that there are theoretical reasons to believe media can reduce prejudice and conflict, but

also plenty to suggest that they cannot, and that no solid evidence exists to resolve this

uncertainty. Next I describe the media program and the setting of the present research—a

reconciliation radio soap opera in Rwanda that began broadcasting a decade after hate

radio helped to fuel Rwanda’s genocide. I explain the theories that motivated the radio

program and the evaluation, and I describe how the program communicated its

educational messages about prejudice, violence, and trauma using a soap opera. I identify

social and phenomenological dimensions of experiencing the program that may play a

role in its impact—such as discussing and reacting emotionally to the program. Then I

describe the yearlong evaluation, a randomized field experiment, and the measurement of

impact. I argue that the pattern of results triangulated across many different types of

outcome measures supports the hypothesis that an entertainment radio serial can shape

social norms and behaviors, even when it fails to convince listeners of its underlying

educational messages. I discuss how this study offers important insights for a future

research agenda on understanding the social psychological mechanisms of media

influence on prejudice and conflict in real world settings.

                 Reducing Prejudice and Intergroup Conflict with Media:
                         Interventions, Theories, and Evidence

       At the turn of the twentieth century, research on the media “redefined the problem

of the media from one of morals, politics, and freedom to one of psychology” (Carey,

1996, p. 30). In 1935 Gordon Allport, who would go on to write psychology’s seminal

book on prejudice, published The Psychology of Radio (with Hadley Cantril), which

explored how radio was “creating a new mental world” (p. 3) and, among other things,

how people used stereotypes when imagining the people behind the on-air voices. Areas

of research on intergroup relations and on the media were growing in the same

environment and frequently overlapped. But some time after the end of World War II,

psychologists for the most part ceded the topic of the media to other disciplines and to

media practitioners.1 The field of communications that grew around the study of mass

media offers many important theories of media influence, but it does not specialize in the

study or measurement of prejudice and conflict as does psychology. Practitioners who

use media to influence prejudice and conflict generally fall into two categories: those

who attempt to incite and those who attempt to reduce prejudice and conflict (for

reviews, European Centre for Conflict Prevention, 2003; Frohardt & Temin, 2007;

Warren, 2005, p. 2).

       Of the two categories of, evidence for media intended to incite prejudice and

conflict stands out. Probably the most infamous example is of the Rwandan radio station

RTLM, which used its yearlong broadcast to goad the population into killing up to

800,000 Tutsi and moderate Hutu civilians over the course of 100 days. The media’s role

in the genocide was deemed so central that the International Criminal Tribunal for

Rwanda charged two founders of the radio station with crimes of genocide, arguing that

the radio set the stage for genocide by broadcasting “ethnic hatred and false propaganda”

with anti-Tutsi songs, jokes, and speeches: “[w]ithout a firearm, machete or any physical

weapon [radio] caused the deaths of thousands of innocent civilians.” (Prosecutor v.

Nahimana, Barayagwiza & Ngeze, 2003, p. 29).

       There are countless examples of efforts to use media for reducing prejudice and

conflict. Each year governments, organizations and corporations pour millions of dollars

into anti-prejudice public service announcements, print and internet publications, and

television and radio programming (European Centre for Conflict Prevention, 2003). But

despite the high stakes of this investment—social, economic, and political—the number

of studies estimating the impact of these media interventions in the world is very small.

Conclusions about the impact of anti-prejudice and conflict reduction media are based on

observational field studies that cannot speak to the causal impact of the program, or on

laboratory studies that change the fundamental nature of the programs by taking them out

of their real world environment and create an unknown bias in translating the measured

processes and effects to the real world.

       Finding out whether the media can reduce prejudice and conflict falls under

psychology’s longstanding and contemporary call to “make psychology matter”

(Campbell, 1969, Miller, 1969; Zimbardo, 2004). Such research would also enrich and

expand current psychological theories of intergroup prejudice and conflict. However

much empirical work lies ahead before researchers interested in media and prejudice can

make this contribution. The most pressing need is for good evidence from real world

settings, which can inform the most basic question—can the media influence prejudice

and conflict? A host of theoretical perspectives predict answers to this question in either

direction, and progress on this complex issue will halt without rigorous empirical

evidence to restart the discussion.

       To accumulate such evidence, psychologists should export their rigorous

experimental methods and measurement techniques into the field. Not only will this

research help begin adjudicating among various theoretical perspectives on how this type

of media can or cannot work, it will illuminate other social and phenomenological factors

that are often invisible to laboratory analogues of media influence, such as the role played

by long term repetition, group watching or listening, discussion of the media, and

emotional and empathic reactions to the media, to name a few.

       Beginning this research first requires an appreciation of the range of media

interventions that exist in the world, of the theories that can be brought to bear on media

influence and intergroup relations, and of the evidence accumulated thus far. I review

these three subjects below.


       Although no official statistics exist, most Americans likely have been exposed to

at least one form of anti-prejudice and anti-conflict media intervention. This exposure

may start with children’s television, like the universally known Sesame Street and other

cartoon and Muppet programs that build tolerance and conflict resolution lessons into

their storylines (see Graves, 1999). Children and adults are exposed to public service

announcements and campaigns against racism, such as the Anti-Defamation League’s

televised “The more you know” campaign,

(, or citywide campaigns on billboards and

bus posters like the Los Angeles “Don’t ruin paradise” and New York “One City”

campaigns (Horovitz, 1993; Internet

sites have joined the anti-prejudice campaign; organizations post educational and

persuasive material against prejudice (e.g.,, bloggers alert

citizens to racism in the news (, and academics

raise consciousness about research findings on prejudice (, Television and film companies release movies

with themes of ethnic and racial understanding (e.g., the Hollywood film “Crash,” now

screened in some diversity seminars; Frazier, 2006), documentaries about hate crimes

(e.g., on a modern day gay lynching Texas; Rojas et al., 2005), and television series that

dramatize historical roots of prejudice (e.g., “Roots,” the story of slave emancipation that

attracted approximately 32 million viewers; Ball-Rokeach, Grube, & Rokeach, 1981).

Some of these media productions are crafted specifically as anti-prejudice interventions;

others anticipate raised consciousness or changed attitudes as secondary effects to the

entertainment value of the program.

       Tolerance and peacebuilding media are also a global phenomenon. Western

countries launch programs similar to those in the US, whether they are anti-racism

billboards in the UK (Vrij & Smith, 1999), advertisements to fight stereotypes of

Aborigines in Australia (Donovan & Leivers, 1993), student newsletters about immigrant

friends in Finland (Liebkind & McAlister, 1999), or a corporate television campaign

against racism in professional soccer during the World Cup (“Stand up, speak out

(against racism)”, Non-western countries are

some of the most active sites of anti-prejudice media interventions. Israel, Palestine, and

South Africa all broadcast their own tolerance-promoting versions of Sesame Street

(respectively, Shara’a Simsim, Rechov Sumsum, and Takalani Sesame, see Brenick, Lee-

Kim, Killen, Fox, Leavitt, & Raviv, in press;, and Macedonia hosts a tolerance-

themed children’s television show (Brussett & Otto, 2004). Liberia and Sierra Leone are

examples of post conflict countries where radio talk shows and dramas aim to promote

ethnic reconciliation (Abdalla & Torrey, 1999; Abdalla, Shepler, & Hussein, 2002). The

primary function of a number of independent media organizations like Internews and

Search for Common Ground is to produce or to support peacebuilding, anti-prejudice

media in troubled societies, (;

         Entertainment-education media is a prominent and widespread approach to media

programming for the social good (Rosin, 2006; Singhal, Cody, Rogers, & Sabido, 2004).

Entertainment-education programs are typically soap operas that weave educational

messages (e.g., about health, literacy, or development strategies) into the plot. The idea

behind entertainment-education media, which draws upon Bandura’s social learning

theory (Bandura, 2002) is that media can “both entertain and educate, in order to increase

audience members’ knowledge about an educational issue, create favorable attitudes,

shift social norms, and change overt behavior” (Singhal & Rogers, 2004, p. 5). Some

examples are a radio drama about Angolan youth from different ethnic groups who play

soccer on the same team (,

and a television soap opera about neighbors and family members coming together to stop

domestic violence (Scheepers et al., 2004). According to listener surveys and

broadcasting statistics, the programs have been remarkably popular (one entertainment-

education literacy program in Mexico boasted ratings that outstripped that of the World

Cup, Singhal et al., 2004). The media program in the present study used an

entertainment-education format.


         The vast amount of time, effort and money invested in media interventions

described above would appear to reflect tried and tested understandings of social

influence and prejudice reduction, or one or more systemic theories of media influence on

prejudice. Even if, as is usually the case, actual interventions are not grounded in theory

(Nzyuko, 1996), the question is whether theories of influence, media effects, or

intergroup relations are useful for predicting their success or failure.

       Psychological theories of social influence (McGuire, 1985; Wood, 2000) have

nourished theories of media effects from the field of communications (see Bryant &

Jennings, 2002). These combined perspectives offer differing hypotheses about the

power and process of media influence, which are of course dependent on context (see

Bryant & Thompson, 2002, and Sood, Menard, & Witte, 2004, for reviews). The

necessary context for predicting the reduction of prejudice and intergroup conflict with

the media could come from some of psychology’s highly developed theories of

intergroup relations.

       However, rather than adjudicating among possible predictions for media influence

on intergroup prejudice and conflict, psychological theories of intergroup relations

complicate predictions. This is in part due to the fact that psychological theories

recognize mass media as contributors to the development and maintenance of prejudice,

but do not theorize beyond this point.2 Mass media figure into psychological perspectives

as carriers of prejudiced or ideological messages, not as a multifaceted phenomena that

warrant study in their own right, to understand how intergroup attitudes or behaviors are

affected by people’s daily interaction with media, their discussions of media with friends

and family, their emotional attachments to media figures and programs, and their reliance

upon media for information and entertainment.

       Scholars reasoning from various respected theories of social influence, media

effects, and intergroup relations could easily disagree about the extent to which media

can reduce intergroup prejudice and conflict. For example, one theoretical approach to

cognitive consistency posits that attitudes toward groups of people can change when the

order of people’s core beliefs or values is destabilized (Rokeach, 1971) and that the

media can destabilize such beliefs (Ball-Rokeach, Grube, & Rokeach, 1984).

Contemporary social cognition research suggest that personal beliefs might change, but

that stereotypes are relatively stable in the long term (Devine & Elliot, 2000), and have

seeped into the unconscious where their operation is routinized and automatic (Banaji,

2001b; Dovidio Kawakami, Johnson, Johnson, & Howard, 1997), which suggests the

potential effects of a media program would be weak. Based on other models of

stereotyping (Rothbart & John, 1985), a media program might reduce stereotypes using a

character who behaves in ways that do not conform to stereotypes about her group, even

though she is otherwise prototypical of that group. Other perspectives would predict that

such media characters may be disliked or discounted if their behavior is too

counterstereotypical (e.g., Costrich, Feinstein, Kidder, Maracek, & Pascale, 1975;

Richeson & Trawalter, 2005). Moreover, media and persuasion theories would point out

that the program must be liked in order to inspire attention, the first step toward influence

(Bryant & Jennings, 2002; McGuire, 1985).

       As another example, norms are thought to have powerful effects on the expression

of prejudice (Crandall, Eshleman, & O’Brien, 2002; Sherif, 1936). While

communications theorists have been impressed by media’s power to change norms

(Mutz, 1998), numerous theoretical perspectives on intergroup relations suggest that

under pressure of social norms, people will veil rather than transform negative affect

towards outgroups (Katz & Hass, 1988; Kinder & Sears, 1981). These theories might

predict that broadcasting new social norms through the media would not reduce the

overall prejudice or conflict in that society. To this point, other theorists might retort that

attitudes follow behavior (Bem, 1972), so if people act in accordance with social norms,

the attitudes will eventually follow.

        To the extent that existing theories make predictions about the impact of the mass

media, these predictions seem to go in many directions. However it is also difficult to

elaborate such predictions because intergroup relations theories do not directly address

the subject of the media. Additionally, the category “media” encompasses many types of

interventions, like radio, television, newspaper and internet, which may call for different

theoretical considerations of their social and phenomenological dimensions. These social

and phenomenological dimensions include factors like watching vs. listening vs. reading,

with another person vs. a group of people vs. alone, discussing the media program with

peers or family, and experiences of empathy or imagination, all of which are

differentially associated with watching television, seeing films, reading the internet, or

hearing a radio drama. Many of these dimensions, such as empathy, are thought to play

an important role in prejudice reduction (Batson et al, 1997; Stephan & Finlay, 1999;

Schecter & Salomon, 2005). Other dimensions, like imagination, might pose novel

theoretical puzzles. As a result, the present study helps to evaluate the utility of existing

models of intergroup relations, social influence, and media effects for understanding the

influence of media on intergroup prejudice and conflict in real world settings.


        There is a large disparity between the number of theories one could bring to bear

on the topic of media influence on intergroup prejudice and conflict, and the amount of

reliable evidence that speaks to this relationship. McGuire’s (1986) classic essay on the

“myth of massive media impact” would apply to the current state of knowledge on the

reduction of intergroup prejudice and conflict using the media. McGuire claimed that

studies of all types of media influence suffer from (a) poor measures of exposure to the

media program, (b) poor measures of the outcome of media exposure, (c) no clear

measurement of a causal relationship between the program and the outcome. While

McGuire questioned the power of media for attitude and especially for behavior change,

he acknowledged that the jury was still out.

       Dividing the existing body of research down the classic fault line of laboratory vs.

field studies3 reveals why very little is known about whether media can reduce prejudice

and intergroup conflict in the world. Laboratory media studies control many factors and

can precisely measure causal impact, but for many reasons their artificiality prevents

extrapolation into the real world. Field studies, on the other hand, illuminate the real

world process of media consumption but are with few exceptions observational and

cannot test for a causal relationship between media and prejudice.

       Laboratory evidence. When anti-prejudice media programs are studied in the lab,

a number of artificial elements prevent extrapolation of findings into the real world.

These elements include experimenter surveillance (e.g., sitting next to and taking notes

on a child while he or she watches an anti-prejudice movie; Houser, 1978), and unusual

settings (e.g., watching television monitors in a converted truck parked on the side of the

road; Vrij et al., 1996). Studies also use media that people would not encounter in the real

world (e.g., reproductions of billboard signs on a series of index cards; Vrij et al., 1999,

and ‘home made’ television commercials that were acted out in slightly different ways

and filmed by the experimenters; Jennings-Walstedt, Geis, & Brown, 1980). Laboratory

experiments also expose participants to very brief excerpts of a media program or image,

and measure impact immediately after (e.g., Bodenhausen et al., 1994).

       Artificiality is particularly damaging in the study of media impact because media

consumption is truly the sum of all its social and phenomenological parts. Most people do

not simply watch a television screen—they simultaneously eat, do homework or chores,

discuss their reactions with a friend, anticipate the reactions of other peers who are

watching elsewhere, and the like. Billboards are often seen while driving and listening to

music; films are watched in theaters where people can hear the laughter or derision of

strangers or observe the boredom of the partner they came with; radio plays in the

background over the course of a day, while people drive alone in their car, or, in less

developed countries, to assembled groups of people. Psychological studies of persuasion

and influence tell us that these simultaneous activities can impact the persuasiveness of,

memory for and interest in the communication (e.g., Janis, Kaye, & Kirschner, 1965).

Thus, the social and phenomenological dimensions of media consumption are extremely

important to understand, but laboratories typically eliminate most of these elements from

the equation (c.f. Iyengar & Kinder, 1987). Whether these factors strengthen, weaken, or

leave media effects untouched is generally unknown.4

       In addition, laboratory studies rarely capture the length of time or the repetition

involved in media exposure. Many media programs unfold over long periods of time.

Drama series, weekly newspaper columns, daily radio talk shows introduce reinforcement

and familiarity effects into the media impact equation, as well as phenomena like loyalty

and emotional attachment. Even if a billboard flashes by on the street or a commercial is

aired for just 15 seconds, a person may see it twice or four times per day. (Too much

repetition may create a backlash against the message; McGuire, 1985, p. 274.)

       There is a subtle difference between using laboratory studies to ask if the media

do reduce prejudice in the world (which demands real-world conditions, if not simply

placing the study in the world), and to ask can media reduce prejudice in the world

(under circumstances rendered in the lab). The typical laboratory circumstances are those

described above, including a captive and concentrated audience, the presence of a person,

the experimenter, who expects prejudice change, and the like. Concerns about the

external validity of the laboratory present the strongest challenges to whether the media

do reduce prejudice in the world. With respect to whether the media can, it is impossible

to say whether laboratory circumstances create a maximal media effect or simply a

different effect. Findings from the laboratory context are generally (and sometimes

suspiciously) positive. One less positive finding, however, is instructive. Anti-racism

billboards designed by a government agency (reproduced on index cards) actually

increased participants’ reported level of prejudice relative to the comparison groups.

Another group viewed a set of reproduced billboards that had been modified according to

various psychological theories of intergroup relations and persuasion. For participants

looking at the modified billboard reproductions there was no reactance—but neither was

there a reduction in racism compared to the control group (Vrij & Smith, 1999).

Exporting these findings from index cards onto real world billboards is questionable, but

nonetheless the experiment is consistent with the idea that psychological theory should be

involved with media intervention efforts.

       Field evidence. The gains in realism brought by moving research into the field are

diminished by the observational design of most field studies, which prevent estimates of

causal impact. Without random assignment, researchers wind up measuring outcomes

among people who choose (or report choosing) to be in the audience, which introduces a

selective exposure bias. For example, it is often noted that people with more egalitarian

attitudes are more likely to consume egalitarian-themed or anti-prejudice media (Ball-

Rokeach et al., 1981; Hur & Robinson, 1978; Sears & Whitney, 1973). Yet controlling

for some of the participants’ observable characteristics cannot solve the interpretational

problem that other unobservable characteristics might be responsible for the measured


       Length of exposure to the program is also often measured poorly in observational

studies; researchers rely on participants’ reports, an untrustworthy practice not only

because of people’s poor recall, but because people may be motivated to misreport. In

poor countries participants might believe that a negative answer will end the interview

and any potential recompense, and in rich countries people might believe a negative

answer will end the interview and set them free to continue on their busy way.

       Field studies of the media and prejudice typically measure outcomes with

questionnaires, which are not always accurate gauges of behavioral change and are

vulnerable to self-presentational biases. (For example, in response to a questionnaire in

Liberia testing the impact of a yearlong program about reconciliation following 12 years

of civil war, one participant said “[the program] made me forget about the war” and

another stated “I have now decided to accept all rebels as my brothers and sisters,”

Abdalla et al., 2002, p. 68). Oftentimes when researchers have omitted a pretest or a

control group, they rely upon participants to report their own change as a result of the

program (Temin, 2001). Most evaluations of public service announcement campaigns

focus on measuring simple awareness of its presence and not behavioral or attitudinal

responses (e.g., Environics Research Group, 2001).

       There have been fewer than ten field experiments worldwide that allow for causal

conclusions about tolerance or peacebuilding media. As a group, these studies capture a

narrow range of media interventions, settings and audiences—all but one are studies of

television programs in North American schools involving younger students (see Paluck &

Green, 2007). Their findings, based on interviews with the students, reveal mixed

positive and null changes. For example, children were more likely to choose pictures of

outgroup children as playmates (Mays, Henderson, Seidman & Steiner, 1975); their self

reported attitudes toward children of other races were in some cases more positive,

depending on the composition of characters in the television program (Graves, 1975); and

children’s moral reasoning about social exclusion based on gender, cultural membership,

and stereotypes improved in most cases (Brenick et al., in press). With respect to the

present research, there has never been a field experiment testing the causal impact of an

entertainment-education program for reducing prejudice or intergroup conflict.


       The present study was motivated by the panoply of media programs taking aim at

intergroup prejudice and conflict, the ambiguous state of theoretical reasoning for how

and why they could succeed, and most of all by the lack of rigorous evidence of the

media’s impact on intergroup prejudice and conflict. The way forward, in my view, is to

collect experimental evidence in real world settings. Prior to theorizing why media does

or does not reduce prejudice and conflict, the field needs an accurate estimate of whether

it does.

           Thus, this research signals a turn away from common practice. The primary goal

is not to test one particular theoretical claim with greater precision, but to provide

rigorous evidence from the real world that can inform broad claims about the impact of the

media. This kind of study can animate future theoretical work by searching for patterns of

effects among a range of cognitive, affective, and behavioral outcomes at the individual and

group level, and by capturing important social and phenomenological aspects of real

world media consumption that suggest mechanisms of the program’s impact. The

multifaceted nature of real world media interventions will complicate but also enrich

theoretical understanding of their effects and will motivate future research that specifies

how the media affects intergroup prejudice and conflict.

           Research like the current study can also expand the boundaries of current

psychological research on intergroup prejudice and conflict in general, outward from its

current focus on milder forms of ingroup preference and discrimination (Brewer, 1999)

and on microfoundations of prejudice like automatic associations and stereotypes (e.g.,

Banaji & Hardin, 1996; Bargh, Chen, & Burrows, 1997). For example, moving to

Rwanda brings topics like violence, trauma, and forgiveness from the margins to the

center of psychological research on intergroup relations (Chirot & Seligman, 2001;

Hewstone, Rubin, & Willis, 2002), and examining the media requires attention to larger

societal forces like social networks, historical narratives, and political context (Latané,

1996; Lustig, 2003).

                                  The Present Research

The context: Rwanda and radio

     Rwanda is an eastern African country slightly smaller than the state of Maryland

and home to approximately 8.5 million people. At the time of this data collection in 2004

and 2005, Rwanda ranked among the least developed countries in the world at 158 of 177

on the Human Development Index (UNDP, 2004).

     Radio is the most important form of mass media in Rwanda (as in the rest of Africa

and the developing world; Bourgault, 1995; Hendy, 2000). In the early 1990’s, as an

intermittent civil war simmered in Rwanda and some ethnic Hutu government officials

were laying the first plans for what would become a genocide, one in every thirteen

Rwandans owned a radio (Chrétien, Dupaquier, Kambanda, & Ngarambe, 1995, p. 57).

Rwandans listen to the radio together in groups, so this figure underestimates the

listening population.

         The 1994 Rwandan civil war and genocide can be traced back at least to a 1959

Hutu revolution that overturned the political hegemony of the ethnic Tutsi minority who

had been propped up by colonial rule. Rwanda’s post-independence Hutu government

continued and extended colonial techniques of ethnic demarcation and discrimination,

this time against Tutsis and also against the 2-3% minority Batwa (pygmy) population.

The 1959 revolution began a cycle in which the uneasy Hutu regime consolidated power

through periodic massacres of Tutsis, keeping an eye southward toward Burundi where a

Tutsi regime massacred members of the Hutu majority (Lemarchand, 1970; Mamdani,


       The anti-Tutsi discrimination and violence in Rwanda, coupled with the growing

Tutsi refugee population in neighboring countries that formed a rebel army, planted seeds

of mistrust in the Rwandan population. In traditional Rwandan society Hutus and Tutsis

lived side by side, spoke the same language, worshiped together, and married one

another. Ethnic identity was often contested and reconstructed by individuals with fake

identity cards or with physical features that enabled members of each group to “pass” as a

member of another ethnic group (called “cheaters,” or abaguze ubwoko). Still, with

encouragement from Hutu political leadership, ethnic stereotypes hardened into

organizing categories of Rwandan politics and, to some extent, ordinary social life. The

historically inaccurate narrative of Tutsis as foreign invaders who had migrated from

Ethiopia and enslaved the indigenous Hutu population justified and perpetuated ethnic

discrimination and offered up Tutsis as scapegoats for the country’s problems. The

Rwandan government maintained its polemic grip on the lives of ordinary Rwandans by

sending policies and communiqués down through Rwanda’s intricate and authoritarian

political bureaucracy, and through the media (see Des Forges, 1999; Mamdani, 2001;

Newbury, 1965; Prunier, 1995).

       The case for the media’s culpability in Rwanda’s 1994 genocide is well known.

The radio station RTLM was launched in 1993 as a music and talk radio station and

progressively worked in anti-Tutsi jokes and propaganda until it was considered an arm

of the extremist government. During the genocide RTLM broadcast rumors that marked

Tutsis and moderate Hutus for assassination and encouraged killers to do their “work.” A

long list of books and essays narrate the crescendo of anti-Tutsi propaganda (e.g. Article

19, 1996; Chalk, 1999; Chrétien, et al., 1995; Gatwa, 1995; Gourevitch, 1998; Melvern,

2004; Thompson, 2007). Radio’s role looms large in journalistic accounts: “When the

radio said it was time to kill people” one journalist wrote, “....the masses slid off a dark

edge into insanity” (in Kellow & Steeves, 1998, p. 124). Over the course of ninety days,

up to 800,000 Tutsi and moderate Hutu were killed by their neighbors and family


       Radio broadcasts were but one of the tools that the Hutu government used to fan

the flames of ethnic division when it was threatened by land shortages, an economic

depression, and a Tutsi rebel army in the late 1980’s (Straus, under review). The more

informed histories of the Rwandan genocide do not attribute ultimate causation to the

media, but rather acknowledge the ICTR’s assertion that RTLM “set the stage” for

genocide by communicating important information and by gradually shifting social

norms regarding Hutu-Tutsi relationships. As one witness stated at the eventual trial of

RTLM following the genocide, RTLM “...spread petrol throughout the country little by

little, so that one day it would be able to set fire to the whole country” (Prosecuter vs.

Nahimana, Barayagwiza, & Ngeze, 2003).

The Rwandan reconciliation radio show: “New Dawn”

       Musekeweya (moo-say-kay-way-ah), or “New Dawn,” is an entertainment-

education radio soap opera, centered on the fictional story of two Rwandan communities.

The story parallels the history of cohabitation and conflict between Tutsis and Hutus over

the course of Rwanda’s history, with each community representing one ethnic group.

Tensions persist because of a history of government favors granted to one community

and not the other, relations between the two communities crumble and the more

prosperous community is attacked, resulting in casualties, traumatization, and refugees.

In an attempt to stop the cycle of violence and repair the communities and their

traumatized residents, some characters band together across community lines, listen to

one another, and speak out against the powerful demagogues from each village who urge

more violence.

       The Rwandan and expatriate staff of a non-governmental organization (NGO)

called LaBenevolencija created the program’s characters and storyline. Rwandan

scriptwriters write the show so as to mirror everyday rural Rwandan society (92% of

Rwandans live in rural areas). The inhabitants of the fictional communities use popular

proverbs and jokes, sing traditional songs, drink banana beer, and in general walk through

the same daily routines of Rwandan life. Most importantly, they wrestle with the same

problems as real Rwandans did in the times leading up to, during, and following the

genocide (LaBenevolencija, 2004).

       Into this plot about everyday Rwandan life and the larger ethnic conflict,

scriptwriters weave “educational” messages about the dynamics of prejudice, the origins

of violence, and trauma healing. These messages are based on the theories of two

American psychologists, Ervin Staub and Laurie Pearlman (Staub, 1989; McCann &

Pearlman, 1990). 5 The messages derived from Staub and Pearlman’s theories read like a

series of declarative beliefs (described below; Staub, Pearlman, & Weiss, in preparation).

The radio characters endorse these messages aloud in a didactic fashion throughout the

yearlong broadcast; typically, characters explain the messages to other characters,

frequently at community meetings or during individual conversations. Of twelve total

messages, one or two are emphasized each episode and are cycled throughout the

yearlong broadcast. The messages are meant to influence listeners’ understanding of and

beliefs about these topics.

       The overall storyline rests on several interlaced subplots, each one following

different characters and emphasizing different communication messages. The most

outstanding subplot is a Romeo and Juliet “star-crossed lovers” story, which follows a

boy from the prosperous (resented) community and a girl from the less prosperous

(aggressor) community. They wish to marry, but their love is forbidden, and when it is

discovered it becomes a lightning rod for tensions between the two villages. Another

storyline involves the girl’s brother, who becomes a vicious demagogue advocating war

between the two communities. His politics dehumanize the members of the prosperous

community, and he eventually goads his comrades into attacking, burning and looting the

community. A third subplot involves the friendship of two elder wise men from each

community, who try to halt the building tensions and ward off the eventual violence. The

two young lovers, in particular the girl, organize the youth of both communities, who also

protest the intercommunity friction and cycle of violence. The comic relief of the show

comes from a subplot starring a “town fool” character—a man who wanders back and

forth between both communities, lampooning the prejudiced and speaking truth to power.

Entertaining and educating: The program’s theoretical mechanisms of change

       The explicit goal of the radio program is to promote understanding of and belief

in its messages about intergroup prejudice, violence prevention, and trauma healing. In

this way, the program’s potential influence is likened to a “public education campaign”

(Staub, Pearlman, & Weiss, in preparation). The program’s fictional storyline portrays

realistic characters and situations from rural Rwanda that may influence listeners in a

more subtle way—through the depiction of desirable intergroup social norms (defined as

socially shared definitions of the way people do behave or should behave; Miller, Monin,

& Prentice, 2000). Thus, program listeners receive explicit lessons on how to understand

prejudice, violence and trauma, and they also receive cues from the fictional characters

about the ways that Rwandans do and should behave so as to reduce prejudice, violence

and trauma. Below I explain the theoretical reasoning behind these two possible effects of

media influence.

       Influencing beliefs. Staub and Pearlman’s twelve educational messages about

intergroup prejudice, violence and trauma healing look like a downsized version of what

previous research named a “belief system” (Converse, 1964; Feldman, 1988; Rokeach,

1968). Converse defined a belief system as a set of beliefs and attitudes that formed a

person’s “...coherent theory or ideology about the social world,” and McClosky & Zaller

(1984) argued that the public absorbs these beliefs through socialization in a particular

political culture. Research that examined influences on the mass public’s belief systems

(e.g., the media, political campaigns) often revealed that people’s belief systems were

rather inconsistent, and also difficult to change (Bem, 1970). Other investigations

suggested that it was possible to “restructure” people’s beliefs with media

communications, and called attention to the power of media to affect people’s “world

view” (Catton, 1960, p. 354).

       Staub and Pearlman (with Weiss, in preparation) suggest that the radio soap opera

can function as a public education campaign, which influences listeners’ set of beliefs

about intergroup prejudice, violence and trauma in two steps. First, characters in the soap

opera teach listeners the twelve educational messages through their “...words and actions,

the consequences of their actions, and the characters feelings about their actions” (Staub,

Pearlman, & Weiss, in preparation). Second, listeners’ understanding of and belief in

these messages will “enable [them] to apply [this] information to their own experience” –

i.e., listeners’ behaviors will stem from their understanding and endorsement of the set of

beliefs promoted by the soap opera.6 Thus, the straightforward prediction is that by the

end of one year, listeners will endorse Staub and Pearlman’s educational messages;

contingent upon their understanding of these messages, they will act upon this new


       Influencing perceived norms. Mass media are thought to be influential purveyors

of social norms, because media depict two important sources of norms: important or

admired figures in society (what they believe and how they act), and public opinion (the

generalized consensus of the audience’s relevant social or political group). Successive

waves of research have ruled that media may not be successful at telling people what to

think, but that the media are very successful at telling people what other people are

thinking (a descriptive norm; see Mutz, 1998; Noelle-Neumann, 1973). With respect to

people’s intergroup attitudes and behaviors, opinions of relevant social or political groups

hold a great deal of sway; according to a classic and contemporary psychological

perspective, conformity is one of the most important causes of prejudice (Allport, 1954;

Crandall & Stangor, 2005).

       Like a real life social group, the soap opera’s fictional but realistic characters may

influence listeners’ perceptions of social norms governing intergroup behavior. Research

shows that audiences often think about fictional media characters as real people (Esslin,

1982), particularly when characters are portrayed as typical people (Shapiro & Chock,

2003), as they are in the present radio program. Audiences form relationships with media

characters that follow many of the same rules of real interpersonal relationships (Rubin &

Perse, 1987). Based on these findings, I propose that the radio characters can influence

listeners’ ideas about how Rwandans do behave (a descriptive norm) and should behave

(prescriptive norms). This particular theoretical perspective is similar to that of role

modeling (Bandura, 1986), but it makes a slightly different prediction. If participants

adopt the social norms portrayed the radio characters, they should not only imitate the

behavior of the characters (a role modeling prediction), but they should also state that this

behavior is normative (that is, it is typical or prescribed.)

        Thus, the second line of theoretical reasoning about the radio soap opera’s

influence is that the program, and in particular, its realistic characters, can communicate

new social norms regarding how Rwandans do behave and how they should behave. The

prediction is that listeners will adopt and behave consistently with the positive social

norms portrayed by the radio characters with respect to intergroup interaction, violence

prevention and trauma healing.

        Belief in Staub and Pearlman’s theoretical messages could change along with

listeners’ perceptions of norms, but it is not necessary to predict that the two effects will

coincide. The literature on social norms shows that people often conform their behavior

to norms even when they do not personally believe in the norm (e.g., Kuran, 1995).

Communicating the messages and norms

        Below I explain Staub and Pearlman’s underlying theoretical principles, and how

they were communicated as set of educational messages and as social norms in the

reconciliation radio program.

       The origins of mass violence can be located in the frustration of basic

psychological needs. One of Staub’s fundamental premises is that the frustration of basic

psychological needs (for security, a positive identity, control, connection to others, and an

understanding of the world) contributes to intergroup conflict. This message is illustrated

in the show when the people of the less prosperous community lose their crops in a heavy

rainfall and experience a famine. Community members hold a meeting where they

express resentment toward and make derogatory comments about the prosperous

community. A wise man stands up and explains to the angry people what their basic

psychological needs are, and how the frustration of these needs tends to make people feel

insecure, frustrated and helpless.

       Scapegoating is a technique for increasing intergroup prejudice and violence.

Closely tied to the message about the frustration of basic needs is a message about

blaming others. Characters like the wise man warn the less prosperous community that

scapegoating is a tactic to transfer their frustrated basic needs onto another group of

people. Another lesson comes in an episode in which the youth form their own

organization against the inter-community violence and discuss the dangers of devaluing

others and blaming them for social problems.

       Mass violence advances along a predictable continuum of events. The overall plot

line running throughout the program portrays Staub’s other fundamental premise that

violence results from a predictable continuum of escalation, from the frustration of basic

needs, to hardened and antagonistic group identities, to hostile behaviors like

scapegoating and small acts of violence, all the way up to intergroup conflict and

traumatization. Some characters explain this point of view when warning others of the

impending clash between the two villages. A narrator who comes on at the end of every

show to summarize that episode’s plot also draws attention to this point.

       These three messages are linked to subsequent, more “proactive” messages about

how to prevent violence, reduce prejudice, and heal from trauma described below. The

proactive messages are the formula for positive characters’ behaviors, which exemplify

new positive social norms.

       Open dissent and active bystanders reduce intergroup prejudice and the

likelihood of violence. The message about dissent and bystanders is that actively speaking

out or acting against prejudice and conflict can help prevent future conflict and

(ultimately) mass violence. Bystanders have a responsibility to be “active”—to speak up

and to slow or halt the escalation of hate speech, discrimination, and violence.

       In the radio program, characters like the young girl from the love story, the wise

men who are friends, and the brother of the extremist demagogue “teach” this message by

announcing to both communities that they should critically evaluate the words and

actions of authority and not blindly obey them. This message is also acted out by some of

the principal characters who band together or act independently against violence, setting

a new norm of active bystandership. For example, the young girl from the love subplot

organizes the coalition youth group that speaks out against elders who plot violence, even

after they are banned from official community meetings. Another man opposes a plan of

revenge launched by the prosperous villagers who were attacked, and even as his

neighbors boo him, he insists on his idea that they should set up a cross-community

meeting. The “town fool” character is a notable personification of this particular

message; he fearlessly ridicules both communities’ bad intentions.

       Intermarriage strengthens ties between groups and reduces intergroup prejudice

and the likelihood of violence. Significant relationships between people belonging to

different groups decreases prejudice and promotes positive relations, according to Staub’s

message about the power of positive intergroup interaction.7 Some of the radio characters

explicitly state their belief in this view, such as in an episode in which the two

communities discuss plans to build a bridge over a river that is too dangerous to cross.

Other storylines and characters function to create a social norm of intergroup interaction

(even if this norm does not extend to all characters in the radio program, which depends

upon renewed conflicts for additional opportunities for learning.) The two wise men from

each village continue their friendship (until one of them is killed), and most importantly

there is the love story between the boy and the girl from opposing communities.

       Trust is important. The importance of trust is not one of the explicit educational

messages, but it is implicit in the general lesson that cooperative intergroup interaction

reduces prejudice and prevents violence. Trust is also a part of the educational messages

about trauma described below, specifically that people should talk about trauma with a

trustworthy person. Characters in the reconciliation radio program point to a lack of trust

as a symptom of the communities’ ongoing conflict, and some of the characters attempt

to promote a new norm of trust through inter-community activities like the coalition

youth group and the collaborative bridge-building project.

       Trauma can be understood, and talking to trusted people is a way to heal from

trauma. With extremely scarce resources for psychological counseling, traumatization

remains a rampant problem in the population (N. Munyandamutsa, personal

communication, January 14, 2005). Psychological professionals in and outside of

Rwanda express concern that Rwandans sometimes interpret trauma symptoms as

insanity or as supernaturally caused (Hagengimana, Hinton, Bird, Pollack, & Pitman,

2003, p. 7). Several communication messages addressed these beliefs, emphasizing that

trauma’s symptoms can be understood, that trauma is not “madness,” and that

traumatized people can heal. These views are expressed by one character who assumes

the role of an informal “trauma counselor” and who speaks to the two communities in

their respective meetings following the attacks on the prosperous village. This character

also educates people about the specific symptoms of trauma, such as isolation, dizziness,

flashbacks, and the like (see Pearlman, 2001).

       The second trauma-related goal of the radio program is to encourage the belief

that trauma can be healed by talking about a traumatic experience with a respectful,

supportive person. This message is also endorsed by the trauma counselor character, and

is established as a new social norm by some characters whose habit is to find trusted

confidantes who listen to their stories with respect and patience.

       Over the course of the yearlong broadcast, each communication message received

focused treatment in at least seven episodes (focused treatment meaning the storyline

revolved around making that communication point and also one or more characters made

that declarative point). The one exception was messages about trauma—these messages

were not as fully developed by the time the evaluation took place (the show continued to

broadcast into a second year, featuring new cycles of violence in the storyline following

the evaluation period so that traumatization could remain a prominent issue.)

       Episodes lasted for twenty minutes, and a new episode was broadcast each week

on Wednesday and Friday evenings, following the nightly news. The show achieved a

high degree of popularity by the end of its first year; a listener survey conducted by the

LaBenevolencija estimated 80% of all Rwandan adults in more populous areas across the

country listened at least occasionally.

Additional factors of program influence

       As previously discussed, many social and phenomenological factors are part of

the experience of media consumption. Two factors are particularly notable in the present

research—the group, in which participants listen and discuss the radio program, and

emotional reaction, particularly empathy, provoked by the soap opera storyline. Group

discussion and empathy are both recognized in the communications literature as

important for media impact generally and in the psychological literature for prejudice

reduction specifically (Katz & Lazarsfeld, 1955; Rojas et al., 2005; Stephan & Finlay,

2000; Zillman, 2006). Although the current research focuses primarily on the main effect

of the radio program, it is also important to document these important factors that are

central to the intervention, which may influence the overall effect.

       Group discussion may amplify the impact of the program. Rwandans listen to the

radio in groups. Radios and batteries are too expensive for the majority of the population,

so Rwandans who do not own working radios gather around the nearest radio at a

neighbor’s house or at the local pub. Gathered together, listeners are much more likely to

engage in discussion and debate about the program during and immediately following the


       Various types of group discussions have been linked to outcomes like

cooperation, political tolerance, reduced conflict, and to more inclusive group identities

(Mendleberg, 2002; Mutz, 2002; Mutz & Martin, 2001). However, other research shows

the group’s initial attitudes may be exaggerated by discussion—thus, discussion among

prejudiced people may simply increase their levels of prejudice (Myers & Bishop, 1970).

Discussion is also integral to some theories of media influence. With respect to prejudice,

Lazarsfeld (1947) and others argued that the mass media are most powerful when they

inspire and reinforce the personal discussions they consider to be necessary for

converting attitudes toward tolerance (Rojas et al., 2005). Bandura’s social learning

theory of the media also emphasizes the importance of discussion and personal

interactions triggered by media consumption (Bandura, 2002).

       Propagandists at Rwanda’s hate radio station RTLM were lay theorists of media

discussion; like many hate mongering programs, their broadcasting style was talk radio

that sounded “…like a conversation among Rwandans who knew each other well and

were relaxing over some banana beer or a bottle of primus in a bar.” (Des Forges, 1999,

p. 70). Whether a broadcast instigates informal discussion among its audience members

or discussion is part of the actual broadcast (as in talk radio), the general prediction from

the literature is that discussion may amplify any effects (positive or negative) of a media


       Empathy for media characters may amplify the impact of the program. Soap

operas—particularly ones that reflect the experiences of their audience, like the present

program—are certain to arouse many emotions, including empathy. Empathy can be

defined as the experience of emotion that is congruent with another person’s situation

(e.g., I feel sad because you are sad; Batson, Fultz, & Schoenrade, 1987).

       Theorists and researchers have linked empathy, in particular imagining the

feelings of other oppressed people, with decreased prejudice (Batson et al, 1997; Finlay

& Stephan, 2000; Schecter & Salomon, 2005). Media researchers propose that the media

can inspire empathy and identification with fictional or real characters (Zillman, 2006).

One way in which empathy for the fictional characters may contribute to reduced

prejudice or conflict is if listeners transfer empathy for characters onto the real life people

the characters represent (e.g., genocide survivors, perpetrators; for a related point see

Andersen, Downey, & Tyler, 2005).

                                        Study Method

       I designed a study that could (a) measure the causal impact of the radio program

(b) in the most naturalistic manner (c) within a representative swath of the population (d)

along meaningful outcomes pertaining to beliefs, perceived social norms, and actual

public behavior. The communities comprising the stratified sample were randomly

assigned to listen to the reconciliation soap opera or to a comparable health soap opera

for one year in their own community settings. At the end of one year, I interviewed

participants individually and in groups, and I observed their behavior with a team of

Rwandan research assistants.

Sampling listeners and communities

       Because radios are relatively expensive to own and supply with batteries,

Rwandans usually listen to the radio in groups. Therefore, I used a group-randomized

design, in which communities are randomly assigned to the treatment (the reconciliation

radio program) or to the control (another radio soap opera about health). My sample of

communities represents the salient political, economic, and ethnic categories of present-

day Rwanda: eight general population communities from the four administrative regions

in Rwanda,8 two genocide survivor communities (mostly composed of Tutsis), two

Batwa communities (the Pygmy minority), and two prisons (see Figure 1).9

Figure 1: Map of research sites

Stars = General population villages; Circles = Survivor communities; Squares = Batwa

       Individual communities, which I will call “clusters,” from each category were

randomly assigned to listen to the reconciliation or health program using a matched

randomization procedure. Each cluster was matched to the most similar cluster from the

same category (general population, survivor, or Batwa) according to a number of

observable characteristics such as the gender ratio, quality of dwellings and education

level. Then one cluster in the pair was randomly assigned to the reconciliation program

and the other to the health program. This stratification of sites helps to balance and to

minimize the observable differences between the communities ex ante. Even after

stratification there could be remaining observable differences, but these were controlled

in ex post analyses of the outcome data. As expected, these controls turned out to have

little effect on my results (described below).

          The final step was to choose individuals to participate within each community.

Official lists of all individuals living in each selected community cluster allowed me to

randomly sample forty adults,10 balancing for sex (where appropriate,11 half male half

female), age (half of the group aged 18-30, the other half above 30), and family (only one

person from an immediate family was allowed to participate). I visited each community

with four Rwandan research assistants, who represented both Hutu and Tutsi (but not

Twa) ethnic backgrounds. The researchers located each individual selected by my

random sampling procedure at home or in the field and explained the purpose and the

procedure of the study. The purpose—“…to understand Rwandans’ opinions about radio

programs and the issues addressed in those programs of the research”—was defined

broadly so as not to create expectations of any particular set of future questions.

          Compliance was nearly 100% at each site. On average one or two selected

individuals at each site turned down the invitation to participate due to poor health or

disinterest. We12 drew replacements from a randomized backup list created in advance.


          When an individual agreed to participate, the researcher conducted a brief pretest

interview to collect information about the individual’s background. Baseline pretest data

for each community cluster served as a randomization check on whether the matched

randomization had divided the sample into two equivalent treatment groups. The pretest

data also served as a control variable to minimize the disturbance terms in the posttest


       The pretest included 11 questions about the participant’s individual background,

including sex, age, occupation, marital status, education, religion, and residential history.

We also asked questions about topics more directly pertinent to the study: radio listening

habits, experience of the genocide (where the participant was, what and whom he or she

lost), and whether any family members were currently in prison.


       The total participant sample was 50% male and the average age was 38.5 years

old. The male-female balance varied across the six pairs of clusters (χ2 = 7.7, p < .05),

because survivor and Batwa communities were heavily female due to the loss of many

men during the genocide, also in two general population villages (Kibuye and Kigali

Ngali). Seventy nine percent of participants identified themselves as farmers, 6% were

teachers or students, and 5% were in business or worked in “town” jobs. The Batwa and

survivor communities were the poorest of all research sites. Seventy three percent of men

and 63% of women in the total sample had some schooling. This figure is different

among the Batwa villages (between 45 and 54% of individuals had some schooling). The

majority of participants in the sample (63%) were married, 22% were single and 13%

were widowed. Catholics made up 64% of the sample, followed by Protestants (14%),

Seventh Day Adventists (9%), “other” Christian denominations (6%) atheists (3%), and

Muslims (2.5%).

Background information related to the study

       Eighty seven percent of Rwandans in the sample reported that they listen to the

radio: 93% of men and 81% of women.13 Only 53% of the participants actually owned a

radio; 83% of those who did not have a radio reported listening in groups with family or


       Ninety-nine percent of the sample was in Rwanda at the start of the genocide, and

approximately 50% were displaced by the violence for a time of one week to a few years.

Sixty-nine percent of the sample claimed one or more of their relatives were killed in

1994; 62% of the participants in general population villages, 100% of survivor and 68%

of Batwa participants. Similar proportions of each community claimed material losses

(property looted or burned) during the genocide. Twenty eight percent of the general

population participants had a relative in prison, compared to 7% of survivor and 57% of

Batwa participants. Table 1 displays a selection of pretest measures that paint a portrait of

the sample, and shows the random assignment as expected produced balanced

covariates—a balance between the reconciliation and health program groups on all

measured characteristics.14

Table 1: Sample pretest measures demonstrating balance between pairs of clusters

                General population communities          Batwa     Survivor
Variable    Kibuye Gisenyi       Kigali-   Butare       Butare-   Kibuye-
                                 Ngali                  KigNig     Butare
Cluster     R H R H R                 H   R     H       R H       R     H
 (% men) 72 58 50 60 56 75 48 33 35 33 20 68
 (mean)    30 44 37 31 37 42 47 41 37 32 38 44
elsewhere 41 23 75 63 73 60 78 63 50 50 48 44
 (% yes)   77 48 78 80 90 54 63 73 45 54 63 80
 radio     92 93 83 80 100 87 75 88 75 85 95 90
Present in
  1994     100 97 100 98 98 100 100 100 100 95 100 100
  relative 69 49 33 36 85 70 73 78 59 76 100 100
 in prison 51 38 25 3 32 23 38 15 50 65 12          3

n           39   40   40    40   41    40   40    40    40   37   41    40

Listening to the treatment (reconciliation) and control (health) programs

       As a treatment, a radio program could be administered in a number of ways,

ranging from rigid and artificial (inviting participants to be observed and questioned after

they listen at a research laboratory) to completely uncontrolled and naturalistic

(broadcasting the radio program and later contacting the population for questioning). I

sought a method that allowed me to be sure that the participants listened to all of the

broadcasts, but without an intrusive monitoring system that would dramatically alter their

ordinary listening experience.

       The same three Rwandan research assistants visited each site over the course of

the year, playing each month’s four 20-minute episodes on a portable stereo for

participants as a group. The participants gathered in a community space as they do for

other non-research occasions to listen to the radio. These visits stretched over the course

of the full year of broadcasting.15 The control group listened to a Rwandan radio program

that addressed a different topic using the same educational-entertainment soap opera

format. Urunana (Ooh-rooh-nah-nah, or “Hand in hand”) is a program about reproductive

health and AIDS that has been popular with men and women in Rwanda since 1999

(hereafter “health program.”) With the control group listening to the health program, the

two experimental conditions were balanced in every respect—all participants had the

experience of following a radio soap opera with a community group over the course of

one year.

       The monthly field visit guaranteed that participants listened to the program but

preserved the most natural environment possible. That the same research assistants

visited every month and socialized with the group helped them to “blend” into what was

designed to be a casual community gathering. They brought a small amount of money to

the site so that the participants could have customary local drinks like urwagwa and

umusururu, 16 as they do normally while listening to the radio in a local cabaret or

neighbor’s home. Research assistants never asked questions about the program or guided

any discussion.

       These “listening visits” overcame one of the primary difficulties of radio and

media program field evaluations, which is that they rely on the self-reported listening or

viewing habits of the research population (McGuire, 1986). I was able to track whether

participants missed a listening session (measuring “failure to treat” rates), and to gather

detailed data about reactions to the program and casual discussions after the program was

finished without putting participants under surveillance in an unnatural environment.

Research assistants filled out observation sheets (see Appendix A) after they left the site

to record attendance and to rate the groups’ levels of observable enthusiasm, interest,

attentiveness, confusion, expressions of emotion, and the amount and type of discussion

during and after the program.17

       With respect to listeners’ discussions, researchers noted how much and what the

participants discussed (independently, without prompting from the researchers) while

listening and also after the program had ended. Of interest was whether and how much

they discussed themes of the program like intergroup prejudice, violence, or trauma, or

features of the program like music or jokes that were unconnected to the messages.

Because the listening experience was designed to remain as natural as possible, these

observations were recorded at a descriptive group level, which does not allow for

correlations with individual listener outcomes.

       That the health program messages do not overlap with the reconciliation program

messages boosts confidence that any observed change pertaining to the reconciliation

program content can be attributed to listening to the reconciliation program. The program

content was effectively the only difference between the two groups’ experiences. The

rest, including listening and observation protocols, research assistants, listening

frequency, and outcome measurement, was the same. Most importantly, random

assignment addresses the chronic confounding variable of personal preference—the

preexisting tastes of the listeners that drive their choice of media programs and their

attitudes and behaviors. Normally, personal preferences obscure the causal relationship

between the media program and their attitudes and behaviors.

Keeping the control group untreated

       We took extra steps to ensure the control group remained “untreated” by the

reconciliation program, which was broadcast during the evaluation period.18 My

necessarily imperfect resolution of this dilemma was to ask the health program groups to

agree to refrain from listening to the reconciliation program for one year. As an incentive

to comply with this promise, they were told that at the end of the year (and the study)

they would receive a portable stereo (a “boom box” with two speakers, two cassette

players, and AM/FM radio) and 14 cassette tapes containing the year’s worth of

reconciliation program episodes (research assistants also promised this gift to the

reconciliation groups). Thus, health participants understood their commitment to refrain

from listening to the reconciliation program as a postponement and not a sacrifice.

       There are several reasons to believe that individuals in the health group made

good on their promise to not listen to the reconciliation program during the study. The

reconciliation program was new; thus there was no pre-existing loyalty to the program.

There were also three other alternative listening choices during Musekeweya’s 20-minute

broadcasting slot. When research assistants made deliberately casual comments to health

participants about the reconciliation program plot halfway through the year, the

participants reassured them that they were not listening. Doubtless some health

participants did listen to the reconciliation program one or more times, because it was a

new show and they were curious, or because they listened to the radio with other people

who controlled the dial. Because participants were unlikely to admit if they did “cross

over,” I have no precise measure of how many listened to the reconciliation program.

       Three aspects of the present research design minimize concerns about various

types of participant non-compliance. First, only the health participants were asked to

refrain from listening to their non-assigned program, because all participants had already

been exposed to earlier versions of the health program on national radio. If some health

participants did not comply with this request and instead listened to the reconciliation

program on their own time, my findings would underestimate the true effect of the

reconciliation program. The true unbiased difference between the reconciliation and

health groups would be larger, given that the reconciliation program influenced some

proportion of health participants as well as reconciliation participants. Thus, the biasing

effect of this type of non-compliance points toward a “false negative” (Type II) error, and

avoids the possibility of a more troubling Type 1 “false positive” error. Second, paired

communities were chosen on the basis that they were geographically close enough to

share many characteristics (so as to balance the treatment and control conditions ex ante),

but they were also far enough apart so that participants from different conditions would

not “treat” one another. Thus, the chances of non-compliance through “participant

interference,” or reconciliation listeners talking to health participants about the program,

are minimal (see Rubin, 1990).19 Third, research assistants measured attendance at each

listening session, allowing for a perfect measure of participant compliance in terms of

showing up to the monthly listening sessions (it turns out that compliance was extremely

high and did not have an effect on the outcome, see below.)

                                        Data Collection


        At the end of one year, a team of 15 Rwandan researchers accompanied the

regular research assistants and the author to each research site for up to 8 hours per day

for three days.20 During this time researchers conducted individual interviews and focus

groups with all 40 participants.21 At the end of the data collection, researchers thanked

the assembled group of participants and gave them the promised portable stereo and set

of reconciliation program cassette tapes (which played a role in an outcome

measurement, described below), and also gave each participant a personal transistor radio

and set of batteries for his or her participation.

        Special considerations. The current Rwandan constitution bans speech about

ethnicity.22 Thus, Tutsis and Hutus were not explicitly mentioned as the residents of the

reconciliation program’s two villages, although the connection between these two ethnic

groups and the two villages was clear to the Rwandan audience. To protect my

researchers and participants during the data collection period, I had to tread carefully

when addressing the topic of ethnicity without obscuring the issues at hand. A

government agency struck the term for ethnicity (amoko) from my questionnaire. I

compromised by substituting the words “types of people” for ethnicity, which was

printed on the official questionnaire carried in the field. Researchers followed up these

questions with an explanation that implied the significance of this term. I am confident

that these questions were understood to implicate ethnicity, because Rwandans are

accustomed to using such “coded” language to refer to ethnicity on a daily basis, and

because it was clear from our participants’ responses that they understood the question, as

many dismissed the coded language altogether and referred directly to Hutus, Tutsis, and


Individual interviews

       We conducted interviews verbally with pictorial response scales to accommodate

the many participants who are not literate. Researchers read participants 37 affirmative or

negatively worded statements. The items were introduced as “statements that other

Rwandans from around the country have made” and the participants were asked to

specify how much they agreed or disagreed with each statement by pointing to one of

four circles printed on a large index card. The circles progressed from small to large; the

researcher explained that the smallest circle represented “disagree strongly,” and the

largest circle “agree strongly.” In the dataset, the circles were coded from 1 (disagree

strongly) to 4 (agree strongly).

Figure 2: Pictorial agreement scale

Ndabihakana      Simbihamya        Birashoboka            Ndabyemeza
 byimazeyo                                                 byimazeyo

From the smallest circle to the largest circle (coded as 1 to 4 in the dataset), the direct
translation of these indicators (explained aloud to participants) are “disagree very
much,” “disagree maybe,” “maybe yes,” “agree very much.” Participants pointed to the
circle that corresponded to their opinion.

Components of the individual interview

       To assess the reconciliation program’s aim of influencing listeners’ beliefs about

intergroup relations and trauma, the individual interview posed a series of statements

measuring participants’ agreement with the communication messages. To test the

hypothesis that the radio program’s presentation could change listeners’ perceptions of

norms about intergroup relations and trauma, the interview posed statements about

descriptive norms (“that is the way things are”) and prescriptive norms (“that is the way

things should be.”)

       To test the discriminant validity of the intervention, I inserted questions that could

test whether the pattern of treatment effects reverses when the effects of the health

program are considered—questions about beliefs and about perceptions of norms

regarding HIV/AIDS, the primary focus of the health program. Questions about empathy

tested the hypothesis that empathy inspired by the fictional drama might transfer onto real

world Rwandans. Behavioral intentions and behaviors relevant to intergroup relations,

intermarriage, and dissent were measured with a combination of interview items and

additional data from focus groups and a separate unobtrusive behavioral measure

(described below).

Beliefs about Staub & Pearlman’s theories of intergroup relations and trauma

        Origins of mass violence. Two statements tested participant’s beliefs about mass

violence vis-à-vis Staub’s descriptive theoretical principles about the continuum of

violence. First, “mass violence grows out of a series of small acts like spreading rumors

and stealing,” and second, “mass violence comes about suddenly.”

        Scapegoating. One statement tested participants’ belief about the roots of

scapegoating: “People who cannot meet their physical and psychological needs are more

likely to blame someone else for their problems.”

        Active bystandership. “If I stand by while others commit evil actions, I am also

responsible” tested participants’ belief in the role of bystanders.

        Intermarriage. One statement tested participants’ belief about the consequence of

intermarriage: “When people marry each other from different regions, religions, or

‘ethnicities,’ 23 this contributes to the peace.”

        Trauma. Three statements about traumatic symptoms and trauma healing tested

participants’ beliefs about trauma—“traumatized people are ‘crazy’,” “perpetrators of

violence can also be traumatized,” and “recovery from trauma is possible.”

       Health. Participants evaluated two statements about AIDS and HIV: “A pregnant

woman who has AIDS can be given a chance to have a healthy baby,” and “You can

safely share something with someone who has AIDS.”

Perceptions of norms about intergroup relations and trauma

       Intermarriage. One statement tested participants’ personal prescription for

intermarriage: “I advise my children (or the ones I will have in the future) that they

should only marry people from the same regional, religious, or ethnic group as our own.”

       Trust. Participants rated how much they thought as a rule “it is naive to trust.” To

anchor this perception of trust norms in the reality of their community, participants also

rated their agreement with the statement “there is mistrust in my community.”

       Open dissent. “If I disagree with something that someone is doing or saying, I

should keep quiet” measured participants’ prescriptions for dissent.

       Talking about trauma. A statement, negatively phrased, tested agreement with

the norm portrayed in the radio program that a person should talk about one’s traumatic

experiences in order to heal: “It is better for my mental health to never talk about the

experiences that have caused me great pain and suffering.” To compare this to self-

reported behavior, participants also indicated whether they had “told someone, like a

friend or a family member, about my experiences that have caused me great pain or


       Health. A final statement testing a prescriptive social norm regarding HIV was:

“It’s necessary that every woman who is pregnant goes to the health center to be tested.”

Other components of the individual interview

        Empathy. To test the prediction that the reconciliation soap opera would

stimulate emotion and particularly empathy toward real life Rwandans, the research team

recorded empathic reactions in two different ways. We measured participants’ empathy

toward real life Rwandans who are represented by the reconciliation program’s characters

by asking whether participants “ever try to imagine the thoughts or feelings of—”

prisoners, genocide survivors, the poor, and political leaders in Rwanda.24 To measure

direct emotional and empathic reactions to the radio programs, throughout the year

research assistants recorded expressions like crying, calling out encouragement, nodding

in agreement, and the like, while participants were listening to the program during the

researchers’ field visits.

        Social distance. The last series of questions asked each participant to imagine a

person who belonged to a “group” that had done harm to him or her in the past. With this

person in mind, participants responded to four classic “social distance” questions: would

they be willing to greet this person on the street, to work with this person, to share a beer

with this person, and would they be willing to allow their child to marry this person?

Focus groups

        The research team pursued certain interview topics further within open-ended

focus group discussions. Participants organized into single-sex groups25 of 10 people

discussed four topics: intermarriage, violence prevention, trauma, and trust. As with the

individual interviews, the goal was to assess participant beliefs and also perceptions of

social norms regarding these topics. Another goal was to see whether individuals were

willing to voice the same opinions in front of their peers as they did in their individual

interviews, an important behavioral measure of dissent.

       Repeating the individual interview item about intermarriage as an open-ended

question, the focus group discussion leader (one research assistant working with a note

taker) asked, “Do you think that when people marry each other from different social

groups this contributes to the peace? If yes, how? If no, why not?” Groups explored the

idea of a continuum of violence with the question “Do you think you could help your

community to avoid more violence? How?” Regarding trauma: “How would you describe

the symptoms of a traumatized person? What would you recommend to someone

according to the symptoms that he or she has?” The question regarding mistrust was also

repeated from the individual interview: “Is there mistrust in your community?” followed

by “Why? How can you rebuild trust?” Due to time and staff constraints, focus groups

were conducted at two of the general population pairs and in the pair of survivor


Behavioral measurement: Deliberations about a communal resource

       Did the reconciliation radio program’s messages have an effect on actual

participant behaviors regarding community issues of consequence? The research team

recorded deliberations that ensued about how to share and supply batteries for the

portable stereo and set of 14 cassette tapes of the radio program that were presented to

each community at the end of the final day of the data collection. Given the monetary and

entertainment value of a portable stereo, this discussion and its outcome was of great

significance to the participants. The other advantage of this measurement is that it

captured spontaneous behavior that participants believed to be “off record”—their

discussions took place during the goodbye party as the research team gathered with all of

the participants to share drinks and to socialize.

       The procedure for observing and recording the deliberations was as follows:

during the goodbye party, the lead research assistant stood at the front of the group and

announced that it was now time to give the group the promised stereo and the cassettes,

which he placed on a bench or table beside him. He followed a set script to invite their

discussion: “you see that there is only one stereo, and it runs on batteries. It belongs to

you, along with all of the cassettes of Musekeweya, and it is up to you to decide how to

use these things, where to keep them when they are not being used, and how to keep the

stereo supplied with batteries. Perhaps, since you are all together now, you can decide the

answers to these questions before you part ways.” Then the research assistant would sit

down. During the discussion that would ensue, two researchers discreetly sitting in the

back of the group recorded the participants’ comments by hand. Of interest was whether

the participants allowed for multiple ideas and viewpoints and whether they expressed

optimism about cooperating to share this community property. I was also interested in

their ultimate decision about how to share the communal stereo.


Analytic issues

       Note from Table 1 that with twelve communities and 40 individuals per

community, the total sample is 480 individuals. However, the “effective N” of this study

is lower than 480, because individuals within each community are not independent

observations. Participants living together in each community share certain unobservable

characteristics that are attributable to their unique shared history and current living

experiences. These “clusters” of people and characteristics violate the independent

observation assumption of regressions used for experiments randomized at an individual

level. Intraclass correlation among members of a community contributes an extra

component of group variance above and beyond individual variance.

       When we do not allow for the possibility that each individual’s disturbance term

is not correlated with the disturbance terms of other individuals in the cluster, there are

increased chances of detecting a significant effect where none exists (a Type I error.)

However, analyzing the clusters as the unit of analysis instead of individuals dramatically

decreases the experiment’s “effective N” (in this case, from 480 to 12), therefore

increasing the chances of detecting a significant difference when one exists (a Type II

error). I adjust the standard errors of the regression to correctly account for the amount of

within cluster co-variance using a robust variance estimator (Arceneaux, 2005). The

robust estimator of the standard errors allows observations to be correlated within clusters

when calculating the overall variance-covariance matrix. With this correction to the

standard errors, I am able to analyze individual level data, rather than aggregating the

data into clusters, which produces the dramatic loss of degrees of freedom and power. I

also increase the power of the experiment by including covariates from the pre-test,

which potentially reduce the intra-cluster correlation among the disturbances and

decrease the disturbance variance.

       In sum, the randomization of the clusters allows for an unbiased estimate of the

program’s effect, and the covariates provided by the pretest boost the experiment’s power

to detect medium to large effects with our sample. Although this experiment may not be

able to detect small shifts as a result of the radio program, we do have the power to test

the powerful effects that are often attributed to the media. Finally, it is important to keep

in mind when considering the sample size that this study is in many ways a

demonstration project that is meant to contribute to what we hope will be a cumulative

enterprise. (See Appendix B for a discussion of the quantitative statistical model and

qualitative data preparation.)

Experimental check

       An average of 35 of 40 participants attended each month’s listening sessions at

each site, with no difference in attendance between listening groups. Concerning

participants’ observed involvement during the listening sessions, there were no

differences between reconciliation vs. health groups’ interest in the program (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), and confusion (m = 1.5, sd = .60; m =

1.5, sd = .78), using the research assistants’ ratings combined across all visits (using a

scale from least to most, 1 to 5). These results indicate that the listening experience was

similar for reconciliation and control, with the key difference being the content of the

media program.

Survey and focus group results

       Individual interviews lasted for an average of 30 minutes, while the focus groups

lasted an average of 1.5 hours. (See Appendix C for a list of individual interview

questions and their frequency distributions.) Investigating the same topics w in individual

and group contexts with quantitative and open-ended response frames allowed me to test

whether observed treatment effects are robust to different measurements26 and to expand

upon precise quantitative findings with descriptive qualitative data. Using questions

posed in both individual interviews and focus groups, I also measured participants’

willingness to speak out about sensitive issues in the presence of their neighbors based on

the difference between their private and public responses.

       Beliefs about mass violence. Two items probing the radio program’s lessons about

violence concerned the idea that violence escalates along a continuum starting with

prejudiced speech and small hostile acts, and that mass violence is therefore not a

“sudden” event. It turns out that the reconciliation program had no effect on listeners’

endorsement of either proposition.

Table 2: Beliefs about violence are unaffected

                         Continuum of            Violence comes
                           violence                 suddenly
                         1       2      3      1      2       3
soap opera              .006   0.04   0.04 0.03 0.01         -.004
SE                      0.10   0.05   0.05 0.12 0.06         0.06
Strata covariatesa       no     yes    yes    no     yes     yes
3 Pretest covariatesb    no     no     yes    no     no      yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       There was no difference between the reconciliation and health groups’ belief that

violence gradually builds along a continuum (M = 3.61, sd = .05, for both groups); there

was also no difference between the groups’ lukewarm endorsement of the idea that

“violence comes suddenly” (M = 2.77, sd = .09).

       In the focus groups, participants from both reconciliation and health conditions

discussed violence as a continuum with remarkable range and insight (after 100% of

participants who spoke stated yes, they could help their community prevent future

violence.) Their collective insights, illustrated with caveats from their own personal life,

fully capture the list of factors identified in Staub’s theory. One participant explained

how violence builds in this way:

          There is poverty, and this makes even small differences great between one person
          and his neighbor, like the poor and the rich. Then there are bad leaders, who
          privilege some to the detriment of others, and this kind of ethnic tension leads to
          contempt of someone who is of a different ethnicity, in conversations and then in
          public speeches in which the other ethnicity is despised. (FG 4-F-1)

These beliefs about violence cannot be attributed to the radio program, because the

control group was just as likely and in some instances more likely to point out factors

along Staub’s posited continuum of violence. In fact, the reconciliation group was more

likely to mention in their open-ended comments a factor that is disputed by Staub’s

theory—“evil people.” Staub emphasizes that average people become violent through

ordinary psychological processes, but as Table 3 demonstrates, a significantly greater

percentage of participants in the reconciliation group (17% of participants) were likely to

cite evil people as one source of violence (compared to 5% of participants in the health


Table 3: Focus group opinions about factors that contribute to violent ideologies in a

Code                                                          Reconciliation      Health

Jealousy / greed / selfishness                                     22.5%          15.6%
Cyclical violence memorialized and perpetuated                     17.1%          11.5%
“Evil”, bad people                                                 17.1%          4.9%
Ethnic or group categorization, discrimination, segregation        14.7%          25.4%
Power politics: divide and rule                                    12.4%          20.5%
Ignorance / lack of education / uncritical mind                     9.3%           4.1%
Poverty                                                             6.2%          10.7%
Family or individual troubles                                       4.0%           4.1%
Rumors and slander                                                  3.1%           3.3%
Restricted political environment                                    1.6%          6.6%
Colonial history of division                                        1.6%            0%
Insecurity                                                          .8%           4.9%
Percentages represent the average number of times each factor was raised in a comment
about violent ideologies out of all comments raised in the focus group discussion.
Percentages add up to more than 100% because a negligible proportion of comments
included more than one factor. Significant differences at the p < .05 level are bolded.

       Beliefs about scapegoating. There was no evidence for changed beliefs about

scapegoating. Both reconciliation (m = 2.78, sd = .07) and health groups (m = 2.80, sd =

.07) on average agreed “somewhat” that the frustration of basic physical and

psychological needs leads people to blame others.

       Beliefs about active bystandership. Participants’ beliefs did not change regarding

a bystander’s responsibility to intervene when others are promoting violence or

intergroup conflict. Reconciliation (m = 3.11, sd = .08) and health groups (m = 3.21, sd =

.07) on average agreed “somewhat” that bystanders share responsibility for what happens

before their eyes.

       This average tendency masks a spike at the disagreement end of the frequency

distribution (see Appendix B), which shows that nearly a quarter of all participants do not

believe that passive bystanders share in the responsibility for any evil acts they witness

and do not try to prevent. Frequently in response to this question, participants recounted a

similar memory of the genocide in which they were unarmed or otherwise helpless to

stop a group of armed people from killing one or more people. These salient personal

experiences and memories may explain some proportion of this trend toward

disagreement with the idea of bystander culpability.

       Beliefs about intermarriage and peace. The intermarriage item probed whether

participants thought that marriage among people from different ethnic, regional, and

religious groups contributes to the peace. Table 4 indicates that the reconciliation radio

program had a modest and statistically insignificant effect in the opposite direction than

predicted, with the reconciliation group being slightly less likely to believe in peace

coming from intermarriage (m = 3.59, sd = .05; m = 3.65, sd = .04).

Table 4: Beliefs about intermarriage bringing peace are unaffected

                        Intermarriage brings
                         1       2       3
soap opera              -0.09   -0.10   -0.12
SE                      0.13    0.10    0.11
Strata covariatesa        no    yes     yes
3 Pretest covariatesb     no     no     yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       Beliefs about trauma. To the exact same degree (m = 1.51, sd = .07),

reconciliation and health groups disagreed that traumatized people are mad. Both groups

believed perpetrators of violence could be traumatized (reconciliation m = 3.29, sd = .06;

health m = 3.45, sd = .05), and that traumatized people can recover (reconciliation m =

3.29, sd = .06; health m = 3.49, sd = .05). However, contrary to the aim of the

reconciliation program, reconciliation listeners were significantly less likely to believe

that traumatized people can recover (see Table 5).

Table 5: Beliefs about trauma symptoms are not affected in the predicted direction

                              Traumatized        Perpetrators can          Recovery is
                                are mad          Be traumatized             possible
                          1       2         3    1      2      3      1        2         3
soap opera              -.001   -.001 -.004     0.06   0.08   0.08   -0.16 -0.15 -0.15
SE                       0.12    0.09   0.06    0.15   0.09   0.09   0.09     0.07   0.08
Strata covariatesa         no     yes   yes     no     yes    yes     no      yes    yes
3 Pretest covariatesb      no     no    yes     no     no     yes     no       no    yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       In the focus groups researchers probed participants’ knowledge about trauma

symptoms. Both reconciliation and health groups listed symptoms like shortness of

breath, social isolation, sudden outbursts, dreams and hallucinations, echoing the

reconciliation program’s portrayal of trauma symptoms. Most participants in both groups

demonstrated they were familiar with the symptoms of trauma through direct observation.

       What, focus group leaders asked, would participants recommend to a person who

has these symptoms? With one exception (discussed in the next section), no differences

emerged between the reconciliation and health groups. Participants discussed how they

could use the social support of the community to heal the traumatized (33% of all

comments). “Visit him many times, make him a good friend of yours” (Gasiza, 14 B-1) is

one exemplary comment. A smaller percentage of participants recommended providing

material support such as sending firewood or a child to help around the house (10%),

visiting a traditional doctor (8%), or going to a hospital (9%). Others mentioned

interventions not recommended by the reconciliation program, such as telling traumatized

people that others suffered more than they or “telling them the reality,” but these

recommendations were found at similar (low) rates in both groups.

       Health beliefs. The overwhelming majority of all participants correctly reported a

belief that pregnant women with AIDS “can be given the chance” to have a healthy baby

(75% of reconciliation and 85% of health, ns). Participants in both conditions also

believed correctly that it is safe to share objects with a person who has AIDS (92% of

reconciliation and 93% of health, ns, see Table 9). Neither result was significant,

although both pointed in the predicted direction that health participants would know more

about AIDS.

Descriptive and prescriptive norms

       Prescriptive norm about intermarriage. With respect to one of the most

noteworthy storylines of the reconciliation program, namely intermarriage, participants

reported whether they tell (or would tell) their children that they must marry within their

own regional, religious, or ethnic group. Here the treatment effect is large and

significant. Those exposed to the reconciliation program are between .25 and .28 probits

less likely to advise in-group marriage. Expressed in percentage-point terms, these

estimates imply that a person who would otherwise have a 50% likelihood of advising in-

group marriage would have roughly a 40% likelihood if assigned to the reconciliation


Table 6: Prescriptive norm about intermarriage is affected

                         No intermarriage
                            in family
                         1      2      3
soap opera              -0.25 -0.27 -0.28
SE                      0.14   0.04   0.04
Strata covariatesa        no   yes    yes
3 Pretest covariatesb     no   no     yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       The finding for shifted prescriptive norms regarding intermarriage is consistent

with data from the participants’ focus group discussions about intermarriage. The

majority of participants agreed with the premise that intermarriage could be a positive

force for peace. Important differences emerged when participants elaborated reasons why

this was so.

       The most common responses did not differ between groups; nearly every group

recounted the same Kinyarwanda proverb: Aho ugishe igisalo ntuhatera ihuye, or “Don’t

throw stones in a place where you keep a treasure” meaning that once a son or daughter

enters into another family through marriage, one must extend goodwill to that family. A

plurality of all comments in reconciliation (39%) and in health groups (36%) echoed this

sentiment, that intermarriage builds a strong social network between families and within

communities. The second most popular reason for intermarriage was the more general

comment that intermarriage reduces intergroup divisions and discord (31% of

reconciliation group comments, 38% health). Participants referred to recent

intermarriages that are flourishing in their own communities and to stories of Tutsis

saved by their association with Hutus during times of violence in Rwanda.

       However, an idea expressed more frequently by the reconciliation focus groups

was that intermarriage sets an example, or creates a new social norm about relationships

between ethnic groups that alters attitudes in the family and the community. For example,

       Sometimes the two fiancés overcome the hate, even when the parents have not.
       But then the [marriage] ceremonies come, and they bring a change of perspective,
       for all those who are invited to come and see them unify. Even if the rancor and
       hate was there, the guests are inspired to reconcile with one another. –(FG #7-11)

Twenty seven percent of all comments in the reconciliation condition focus groups

stressed the utility of intermarriage as a public act that can shift social norms about

intergroup relations, compared to 5.7% of all comments in the control group. This

difference is statistically significant using a probit regression on the numerical codes of

comments, β = .21, se = .08 p < .01.

       By contrast, health focus groups were more likely to describe intermarriage as a

private choice and challenge, rather than in terms of a decision that would involve and

potentially transform their social and familial environment. For example,

        “...some youth nowadays have dispensed of these ideas of division between
       ethnicities. Once the child loves a person, you could not, even if you were the
       father, convince the child to leave that person because of ethnicity. In that case,
       intermarriage overcomes ethnicity.” (B-1, #13-5)

Health condition groups supplied this kind of answer 11% of the time; reconciliation

groups did not mention this idea.

       Both reconciliation and health groups placed conditions on the positive effects of

intermarriage, lending credibility and realism to their otherwise positive discussions.

Many cautioned that positive effects depend on the “gravity of the [political or social]

situation,” citing stories about Tutsis who were killed by Hutu spouses or in-laws during

the genocide. A small minority of participants in each group stated that intermarriage was

against social norms , e.g.,: “Many people think it should not happen, you can remember

one of our leaders who forced his daughter to have an abortion when she was pregnant

with a Tutsi’s baby.” (FG discussion #M-1 11).

       Norms about trust. Reconciliation group members were significantly more likely

to deny “it is naïve to trust people,” at a level of “strongly” disagree vs. “somewhat”

disagree, β = -.20, se = .10 p < .05. Reconciliation groups disagreed on average 1.81 (sd

= .07) and health groups on average 2.01 (sd = .08).

       Reported trust: alone and in focus groups. Responses to the individual interview

item “there is mistrust in my community” were on average affirmative—participants in

both responded “somewhat agree,” a 3 out of 4 on the rating scale (health mean = 3.1, sd

= .07; reconciliation mean = 3.0, sd = .07; see Table X).

       However, in focus groups, when participants were asked in front of fellow

community members about the level of trust in their community, some responses shifted.

Notably, it was health group participants who were most likely to modify their private

stance about mistrust in their community. Thirty nine percent of the health groups’

comments were unqualified denials of mistrust, compared to 7% in the reconciliation

focus groups. Taken in light of the uniformly high levels of mistrust reported in the

individual interviews, the difference in focus group responses seems to reveal more about

the reconciliation group’s willingness to speak out on difficult subjects than about actual

levels of community mistrust.27

       Norm about open dissent. Survey participants’ evaluation of the statement “If I

disagree with something that someone is doing or saying, I should keep quiet” revealed

one of the strongest treatment effects associated with the reconciliation program. Those

exposed to the reconciliation program were .26 to .29 probits less likely to endorse this

statement (see Table 7).

Table 7: Norms about trust and dissent in the community are affected

                                 Naive                   There is                 I should
                                to trust                  trust                   Dissent
                        1          2        3      1        2        3     1         2     3
Reconciliation                                                              -
soap opera            -0.16 -0.15 -0.14 -0.16             -0.13     -0.10 0.26     -0.28   -0.29
SE                    0.08       0.07      0.07   0.18     0.10     0.07   0.18    0.10    0.07
Strata covariatesa         No      yes      yes     no       yes     yes   no       yes    yes
3 Pretest
covariatesb                No       no      yes     no       no      yes   no       yes    yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       Norm of talking about personal trauma. Reconciliation participants were much

more likely to agree that people should talk about traumatic experiences, an effect of .17

to .22 probits (See Table 8 about trauma below). This normative position did not correlate

with a higher rate of self-reported talking, most likely because 83% of people across both

groups report that they had already talked with someone about their traumatic


Table 8: Norm of talking about trauma is affected

                           Talking about
                          1       2     3
Radio Program           -0.17   -0.20 -0.22
SE                       0.08   0.02   0.03
Strata covariatesa         no    yes   yes
3 Pretest covariatesb      no    no    yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       Consistent with the findings from the individual interviews for talking about

trauma is the observed difference in the focus group discussions of trauma healing.

Reconciliation listeners mentioned the importance of talking about trauma and listening

to other people talk much more often than health groups, β = .15, se = .07. For example,

       “You should accept his condition and let him express his state of mind.” (G-G, 10

       “The most important thing is to accept all that she is. After that, approach her and
       listen to her attentively without wounding or rushing her.” (R-B, 11, M)

       Norm about health.       Although general agreement was extremely high (95%),

there was a small statistically significant tendency for health groups to agree more often

that all pregnant women should be tested for disease (see Table 9). Predicting outcomes

using the .56 probit coefficient shows that listening to the health program made

individuals 1% more likely to state that pregnant women should be tested for AIDS. This

tiny probability was affected by the very high rate of agreement in both groups.

Table 9: Beliefs are unaffected and norms are affected by the health soap opera

                        Belief: pregnancy      Belief: share with          Norm: test
                            and AIDS            AIDS patients           pregnant women
                         1       2     3       1        2      3        1      2      3
program                0.02    0.06    0.06   0.04    0.07    0.10    -0.41   -0.43   -0.56
SE                     0.16    0.05    0.05   0.21    0.14    0.16    0.22    0.20     0.18
Strata covariatesa      no     yes     yes     no      yes     yes     no      yes     Yes
3 Pretest covariatesb no        no     yes     no      no      yes     no      no      Yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

Social distance measures

       I constructed an additive index of social distance, in which the target was a person

who had in the past harmed the participant or the participant’s group. The index

represents how much participants agreed that they would greet, work with, share a beer

with, or marry their child to such a person. Participants’ willingness to interact was at a

relatively high level: somewhat agree for reconciliation groups (m = 3.07, sd = .06) and

for health groups (m = 3.11, sd = .06). However there was no difference, substantively or

statistically, between the two groups’ willingness to be close to that person (this result

held for each individual question). The reconciliation program’s strong themes of cross-

group friendship and forgiveness were not reflected in this measure of listeners’

orientation toward reconciliation in their personal social lives.

Table 10: Desire for social distance from “those who have harmed you or your group” is

                          Willingness to
                        greet, work, share,
                         1       2      3
soap opera              -0.02 -0.04 -0.04
SE                      0.18   0.10    0.11
Strata covariatesa       no    yes     yes
3 Pretest covariatesb    no     no     yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       Behavior in communal resource deliberation. Differences appeared marked in the

general character and length of deliberations within the reconciliation and health groups

about the communal stereo and tape cassettes even before systematic coding of the

discussion transcripts. Reconciliation group deliberations seemed livelier and more

contested, and participants spoke positively about their ultimate ability to collectively

manage stereo use. Within the health groups, researchers frequently observed the

following pattern: the first member of the group to offer a comment would propose

handing over the stereo and cassettes to the village’s local authority, who could regulate

usage and financial contributions for the batteries. Following this proposal, participants

would overwhelmingly support the motion and close the matter.

       In the reconciliation groups, deliberations frequently followed a different pattern.

After the same initial proposal to entrust the stereo to the authorities, one or more of the

participants would challenge this suggestion, claiming that the group should be

collectively responsible for the stereo, or that the group should elect one of their members

to manage usage of the cassettes. Positive comments about the group’s ability to

cooperate are illustrated by this participant in a southern reconciliation village: “We’ve

been coming together to listen all of this time, why can’t we come together to listen to

this stereo together, just as we did before?” (Site S-B).28 In contrast, the health group that

was statistically matched to this village had so many problems organizing themselves that

they suggested the research team should take it back, because it would cause too many

problems (Site S-N).

       These different patterns were borne out by statistical analyses of the coded

transcripts. To measure whether participants raised and debated a diversity of opinions

about how to handle this communal good, I counted the number of solutions raised and

the number of dissenting statements in each group. I combined these two counts to create

an index that represents a balanced deliberation session, in which participants propose

and discuss various suggestions. A comparison of this index reveals that the

reconciliation groups discussed this communal dilemma longer and debated more

viewpoints on how to share it than health groups (χ2 = 4.67, p < .05). I also counted the

number of positive comments groups made about their ability to cooperate on this

communal task: reconciliation groups made more comments like this, at a level of

borderline significance (χ2 = 5.67, p < .06.)

       In the end, nearly all of the reconciliation and health groups decided to share the

stereo and cassettes collectively and to listen together as a group.29 However, the process

by which they arrived at that decision, and their anticipations of the groups’ ability to

work together were consistent with the emerging pattern that reconciliation participants

became more open to dissent and to community interaction and cooperation.

       Empathic reactions. According to the monthly field notes, participants’

emotional reactions to both programs were visible, audible, and frequent. There was not

one listening session in which researchers did not document more than one reaction of

surprise, happiness, scorn, anger, or sadness, and typically a wide combination of these

reactions occurred. Some of the numerous examples of emotional reactions are:

participants crying out in pain when a man from the prosperous community was beaten

(Note 78, KibGPT), tears in the eyes of men and women participants during the funeral

of one of the two elderly friends from opposing communities (Note 20, ButGPT),

laughter and excited clapping during a reunion of the star-crossed lovers (Note 204,

KNGPT), and calls of encouragement to the girl when the relationship was foiled again—

“ihangane sha” (hold on dear; Note 104). The health program also inspired strong

emotional reactions, including included anger and shouting about an unfaithful lover, and

sadness and wailing for a victim of AIDS.

       As described above, some emotional reactions were expressions of sympathy

(feeling sorry for a character) and others were of empathy (feeling an emotion parallel to

the character), although these differences are not always entirely clear. In the individual

interview, an additive index of questions about empathy (“do you imagine the thoughts or

feelings of”) measured empathy for categories of real life Rwandans who are portrayed

by the fictional characters in the reconciliation program (prisoners, genocide survivors,

the poor, political leaders). There was a moderate and significant effect such that

participants exposed to the reconciliation program expressed more empathy across all

four categories (See Table 11).30

Table 11: Increased empathy for real life Rwandans who were portrayed in the radio soap

                         1   2     3
soap opera              0.16   0.17   0.17
SE                      0.11   0.09   0.08
Strata covariatesa       no    yes     yes
3 Pretest covariatesb    no     no     yes

Entries are ordered probit estimates of the treatment effect. Standard errors are adjusted
for clustering, except in cases where the estimated clustered standard error is smaller
than the conventional standard error. Significant probit coefficients are bolded.
  Type of community: general population vs. Batwa community vs. survivor community
  Sex, birthplace, regularity of radio listening

       Discussion. The field notes reveal similar amounts of spontaneous participant

discussion during and after both radio programs, supporting the claim that participants

experienced comparable listening environments in the two experimental conditions (with

the exception of program content). The levels of discussion are high, although not so high

as to prevent attentive listening. On a scale from 1 (completely silent) to 5 (constant

commentary) researchers rated their levels of discussion during the broadcast on average

3.09 (sd = 1.08) across all monthly visits to groups in either treatment condition. After

the broadcast, researchers estimated on average that participants spent 63% of their time

discussing the program together before parting (sd = 25.0), again with a negligible

percentage point difference between reconciliation and health groups.

       What were these discussions like? Field notes reveal that participants kept up a

running commentary as they listened, on the actions and conversations of the radio

characters. Listeners echoed, supported, or protested their statements, predicted how

other characters might react, and the like. Participant noises, whistles, and exclamations

(“eh!” “yoo!”) frequently punctuated the dialogue and music of the programs.

       Participants also discussed the overarching messages of the program, not simply

plot developments and jokes. For example, one episode of the reconciliation program

ended with a character’s comment that tolerance and respect for one another’s ideas is

necessary, to which a male participant called out “we should repeat those words!” (M

Kibuye MRT), thereby sparking a discussion. I noted a similar pattern in the health

groups, whose chatter was always sparked by storyline lessons about sexuality or HIV.

Local proverbs used to illustrate a point also inspired dialogue, in which participants

traded likeminded proverbs of their own.


       The yearlong reconciliation radio soap opera influenced listeners’ perceptions of

social norms and behaviors regarding intermarriage, open dissent, trust, and talking about

personal trauma. The radio program did not affect listeners’ beliefs regarding the

program’s educational messages about the etiology and dynamics of intergroup prejudice,

violence and trauma. This pattern of results was consistent in the comparison

experimental condition; listening to a health radio soap opera influenced perceptions of a

norm about getting testing for AIDS but not beliefs about AIDS. Both radio programs

aroused emotions and expressions of empathy for the radio characters as participants

listened, but reconciliation listeners expressed higher levels of empathy for real life

Rwandans who were portrayed in the reconciliation program. Listening to the soap

operas in groups facilitated spontaneous discussions about the program.

       More specifically, the educational messages built into the plot of the

reconciliation program failed to influence listeners’ beliefs about the etiology of mass

violence (as one end of a gradual buildup), causes of increased prejudice and violence

(scapegoating and passive bystanders), and trauma (symptoms and potential for

recovery). In some cases the predicted effect of the educational messages went in the

opposite direction—reconciliation program listeners were more likely than health

program listeners to believe that “evil people” are among the primary causes of violence,

less likely to believe intermarriage can bring peace, and less likely to believe people can

recover from trauma.

       In contrast, I found that the reconciliation radio program influenced listeners’

perceptions of social norms and their subsequent behaviors. Reconciliation listeners

reported that they were less likely to proscribe interethnic or other types of outgroup

marriage for their children and they were more likely to describe intermarriage as a

public act that can change community norms about intergroup relations. Although

mistrust was just as common in reconciliation listeners’ communities, they were less

likely to view trust as “naïve.” Prescriptive norms to speak out about difficult topics were

most notably affected by the reconciliation program. Listeners claimed they felt they

should speak out when they do not agree with their peers, and that they should talk about

their personal trauma in order to recover from the experience. Their behaviors backed up

these claims: reconciliation listeners were 32 percentage-points more likely to speak up

about the difficult subject of community mistrust in front of their neighbors. In addition,

during group deliberations about how to manage a communal resource, reconciliation

listeners were more likely to engage in an open and balanced debate by proposing a

diversity of suggestions and dissenting opinions. Reconciliation listeners expressed more

optimism about cooperating with group members to share the resource (although this

finding was of borderline significance). Undoubtedly, listening to and discussing the

radio program in a group had an important impact on these results, which I discuss more

below. Taken together, the evidence suggests that changes in perceived norms about

intergroup interaction, trust and dissent proceeded hand in hand with altered listener


       That listeners’ intergroup behavior and their perceptions of intergroup norms

changed despite no change in their personal beliefs is consistent with a deep social

scientific literature on conformity, public opinion, and intergroup prejudice (Allport,

1954; Asch, 1958; Festinger, 1957, Kuran, 1995; Miller, Monin, & Prentice, 2000;

Sherif, 1936). This literature reveals a human tendency toward preference falsification

(Kuran, 1995), or the tendency to conceal one’s personal beliefs in favor of conforming

to a perceived social consensus. Research on social norms and intergroup relations

reveals that norms have a powerful influence on individuals’ expressions of tolerance or

of prejudice, regardless of their “genuine” views (Crandall et al., 2003). This evidence

supports the longstanding idea that “[m]uch prejudiced behavior does not stem from

prejudiced attitudes or motives, nor even from faulty information, but rather from the

need to conform to prevailing social norms” (Chein, 1947, p. 415). In the present

research, it was not necessary for listeners to endorse the radio program’s underlying

messages about positive intergroup relations for them to behave in a more open and

cooperative manner with their peers, or for them to report intergroup norms that

supported trust and intermarriage.

       The disparity between findings for personal beliefs and those for perceived norms

and behavior also fits with a growing consensus on the dynamics of violence during the

Rwandan genocide. Scholars emphasize that violence became normative and did not

necessarily reflect the personal prejudices of the killers (Fujii, 2006; Straus, 2006). The

“norm of violence” was communicated by authorities, neighbors, and by the media; these

various sources of social pressure compelled citizens to kill (Straus, 2006). The sobering

lesson is that normative pressure, through the media and other sources, can be a double-

edged sword.


       The prediction at the outset of this research was that like real life peers, the radio

soap operas’ characters would influence listeners’ perceptions of social norms and their

subsequent behaviors. This prediction was based on research showing that peer opinions

and public opinion are both powerful influences on an individual’s attitudes and

behaviors (Blanchard et al., 1991; Sherif, 1936, Mutz, 1998), together with findings that

fictional media characters are sometimes thought of as real people (Blumler & Katz,

1974; Esslin, 1982) and that more generally, fictional stories can influence real world

attitudes (see Gerrig, 1993; Prentice & Gerrig, 1999).

       While the present research did not directly measure listeners’ perceptions of the

soap opera characters, a great deal of indirect evidence suggests that participants came to

see them as realistic figures who were directly relevant to their lives. Field notes show

listeners reacting strongly to the good and bad predicaments of the characters, gossiping

about the characters with other listeners, and sometimes directly addressing characters

while the show was playing (e.g., “Don’t say that, Shema!” to the male star-crossed

lover, when he announced he had given up hope for the relationship, KI-GP 32). In

addition, reconciliation listeners in several sites reported that they had nicknamed other

people in the community using the names of certain soap opera characters—pretty girls

were named after the female star-crossed lover, respected old men were named after one

of the show’s two wise men. More direct evidence of the realism of the show’s characters

comes from a separate survey of people listening to the reconciliation program, in which

95% of respondents said that the reconciliation program’s characters reminded them of

people from their local community (Paluck, 2006). Together, these pieces of evidence

indicate that characters were perceived as realistic and typical of the types of people in

listeners’ environment. Overall I believe these findings support the idea that radio

characters were able to influence social norms and behaviors because of their perceived

similarity and relevance to the listeners’ life experiences (see also Bandura, 2004).

       I offer a similar explanation for the finding that the radio program increased

empathy for real life Rwandans—prisoners, genocide survivors, the poor, and leaders.

Positive feelings about individuals may transfer onto the groups they are identified with

(see Andersen, Downey, & Tyler, 2005). By this logic, listeners’ emotional, empathic

reactions to the soap opera characters who were imprisoned, survivors of violence, poor,

and in leadership positions may have transferred onto the real life versions of these

groups because the soap opera characters were considered to be representatives of these

real groups. These explanations for the transfer of norms and empathy from the fictional

world to the real world merit further testing, but for now the evidence does seem to

support other research showing that media audiences do not draw a bright line between

the fictional and the real world (Harris, 2004, pp. 49-51; Gerrig, 1993).

       As I indicated above, the effects of the radio soap opera are inseparable from the

effects of listening to the soap opera in a group. Individual listeners became aware of

social norms portrayed in the radio programs, but in groups they also became aware of

other people’s awareness of these norms. Moreover, if group members react positively to

these norms, their endorsement creates another vector of social influence on the

individual listeners (Katz & Lazarsfeld, 1955). Field notes documenting the groups’

listening sessions are rife with examples of how the group creates an echo chamber in

which a message coming through the radio ricochets through the group, hitting each

member many more times than if he or she had been listening alone. Group commentary

about a message was as often evaluative as it was repetitive. Participants called out in

agreement, nudged one another, recited proverbs consistent with the character’s actions,

and even moralized—“we should repeat those words!” Positive endorsement of the

messages turned perceived prescriptive norms into actual group norms: when the group

voiced agreement with a certain message about how one should behave, it became

apparent to all group members that, at least when they were in that group, that is how

they actually should behave. For example, when participants spoke their minds about

how to share the communal resource in the outcome evaluation, they probably did so

with the expectation that their fellow group members shared their perception of open

dissent as a positive behavior.

       Perhaps even more so than the ongoing commentary, group discussion that arose

after the radio episodes ended also created opportunities for persuasion and for

cultivating shared understandings of norms. Group discussion of a persuasive message

can enhance its persuasive effects (Lewin, 1952), sometimes because of social pressure

from those who agree with the message (Mendleberg, 2002), and other times because the

discussion generates more thoughtful processing of the message (Petty & Cacioppo,

1986; for all of the ways discussion can persuade or dissuade, see Mendelberg, 2002;

Wood, 2000). In general, discussion of the radio soap opera, which often focused on the

program’s substantive messages, contributed further to what psychologists call “socially

shared cognition” (Fiske, 2005, p. 44), or a collective awareness and understanding of

radio program. Collective understandings are the building blocks of social norms.

       The group aspect of the intervention may have been just as important as the radio

program itself for the effect sizes that were found. It has long been thought that

interventions (like media programs) that target groups or communities should be

particularly effective for reducing intergroup prejudice and conflict. If one tries to change

an individual’s prejudiced attitudes with individual education, Allport (1954) asserted,

“[t]he chances are this lesson will be smothered by the more embracing norms of his [sic]

family, gang, or neighborhood.” Peers and opinion leaders are often the key link between

a media message and an individual’s adoption of the message (Katz & Lazarsfeld, 1955;

Rogers, 1995); moreover as discussed above, groups give individuals social permission to

express and to act upon a message when it is clear that the group is aware and approves

of the message. Careful study of the joint influence of media messages and peer and

community groups on prejudice will be a useful path for future research.

       Certain effects of the reconciliation program were notably large relative to others,

and it is important to ask why this was the case. The biggest shifts occurred in

reconciliation listeners’ perceptions of social norms and behavior vis-à-vis open dissent

and intergroup relationships (intermarriage and cooperation). Both subjects have

particular resonance in Rwandan culture, which may have made the ground particularly

fertile for influencing listeners’ relevant perceptions and behaviors.

       Specifically, there exists a prominent imperative against dissent in Rwanda’s

hierarchical society. Proverbs like “when you are in a weak position, rest calmly and

cross your arms” illustrate the customary prescription to refrain from disagreeing with

authority (this proverb was actually used in the reconciliation program to make this

point.) Rwandans frequently note and critique this cultural norm: “Rwandese are like

cows, we go where we are told” one of our participants noted with some disgust, an

observation that was repeated many times in our interviews. Even when speaking with

peers, Rwandans value indirect, sometimes circuitous ways of communication as opposed

to direct disagreement (Overdulve, 1997). Likewise, but in a different vein, cooperation

and interdependence among neighbors are salient and celebrated ideas about customary

Rwandan society. Rwandans note with pride the various imperatives to share beer with

neighbors from the same straw, to carry ailing neighbors down the mountainside on an

ingobyi (stretcher), and other rituals that characterize peaceful Rwandan life. When

participants were asked to define what reconciliation means in the pretest interviews,

their responses were remarkably uniform such that the large majority included these

rituals of interdependence and cooperation like sharing beer and carrying the sick

(Paluck, Green, & Nzamukwereka, 2004).

       Reconciliation listeners’ relaxed stance on the issue of intermarriage is

particularly impressive. Previous studies of prejudice have found that attitudes toward

intergroup marriage are among the most reliable indicators of out-group hostility

(e.g.,Green, Abelson, and Garnett, 2000). In Rwanda, views of intermarriage have also

served as a kind of weathervane for the state of Hutu-Tutsi relations. Intermarriage prior

to the 1990’s was relatively common, although the exact rates are unknown (Des Forges,

1999). In the early 1990s, the government used the issue of intermarriage as a way to

promote its discriminatory and then genocidal ideology (e.g., by prohibiting army officers

from marrying Tutsi wives; Verwimp, 2004). After witnessing the nihilism of a genocide

in which some Hutu men killed their Tutsi wives, current pessimistic views of

intermarriage are understandable. But many Rwandans have also witnessed some of the

positive effects of intermarriage; an investigation into the lives of Rwanda’s intwalis (“the

righteous ones”) who saved Tutsis during the genocide found that they were more likely

than not to be related to Tutsis through intermarriage (PRI, 2004). The reconciliation

soap opera’s love story between the girl and the boy from opposite “communities”

represented a very important pre-existing narrative about Hutu and Tutsi relations in

Rwandan culture.

       Thus, one straightforward way to interpret these strong reconciliation program

effects is that attempts to shift social norms may be most successful when they give force

to an idea that already has momentum in a society’s culture or history or in the private

views of its population.31 The program’s love story may have reminded listeners of the

formerly positive status of intermarriage in their culture, or of stories about the lifesaving

effects of intermarriage during the genocide; scenes featuring radio characters speaking

their minds may have inspired listeners who feel unhappy with Rwanda’s

nonconfrontational standard.

       Finally, why did the radio program fail to achieve its goal of informing and

changing beliefs about its underlying theoretical principles regarding intergroup

prejudice, violence and trauma? This study was not the first to find that beliefs did not

change from media exposure, although it was one of a handful of studies to measure this

relationship in a field experiment (Ball-Rokeach et al., 1984; Oskamp, 1989; Schofield &

Pavelchak, 1989). Perhaps there was a null effect because, as others have observed,

beliefs are notoriously difficult to change (Bem, 1970; McGuire, 1986; Wood, 2000) and

media are not be good at “telling people what to think” (see Kinder, 1998; Mutz, 2002).

However in a few instances in the present research beliefs did change—in the opposite

direction than predicted.

       The non-findings with respect to beliefs may also be explained by the notion that

there is particular resistance to messages aimed at beliefs about important personal

experiences. Research demonstrates resistance to persuasive messages when people’s

personal values or their “egos” are involved (Johnson & Eagly, 1989; Sherif & Hovland,

1961). Other investigations have demonstrated that people are less likely to be influenced

by fictional stories when the stories overlap substantially with their real lives; in these

cases, people process the stories much more critically, and dismiss information that is not

consistent with their own experience (Prentice, Gerrig & Bailis, 1997). In the current

study, the reconciliation program attempted to influence the belief that violence does not

comes suddenly. Many Rwandan peasants, who observed a buildup of tension among

Hutus and Tutsis and among political parties prior to the genocide, still claim they

experienced the genocidal violence as a great surprise. Many participants made

comments similar to this one during the interviews: “When the violence began, it fell

upon us like a sudden rain.” (Kig 14B).

Future questions

       The findings from this field experiment tell a more open-ended story relative to

laboratory experiments that strive to isolate a single independent variable, and one or

more variables that identify a process of change. While I can state that the total

intervention had a causal impact on norms and behaviors related to intergroup prejudice,

the intervention consisted of many interrelated factors that may have contributed to this

finding and that animated various processes of change. The intervention included a

reconciliation soap opera, involving an entertainment program featuring humor, fictional

drama, realistic characters and educational messages; the phenomenological experience

of radio listening, which involves the auditory system and imagination; and the social

element of listening in groups, which involves perceiving other people’s reactions,

comments and participating in subsequent discussions about the program.

       To paraphrase Kurt Lewin, in order to truly understand a system you must change

it (Shein, 1988). Manipulating different factors of a radio intervention in future field

experimental research will enrich and expand understandings of why, how, and when the

media exerts an influence on intergroup prejudice and conflict. With experimental

methods, one can systematically manipulate variables like humorous content, discussion

(e.g., whether it is encouraged or not), and modality (e.g., radio vs. television) to measure

the relative importance of each factor in different contexts.

       Below I discuss some notable aspects of the current radio intervention that may be

important for understanding the current findings and media influence on prejudice and

conflict more generally: entertainment (including humor and fictional narrative), social

experiences of media (like the presence of others and discussion), and listening

(involving mental imagery and imagination). Finally, I note how the real world setting of

this experiment broadens the conception of intergroup prejudice to include related facets

of intergroup relations like violence, trauma, and historical narrative.

       The role of entertainment, humor, and fiction. How important was it to the present

findings that the radio intervention took the form of an entertaining fictional drama?

Would an informative radio documentary or a series of speeches have similar effects?

Two intuitive predictions might be that entertainment has a stronger influence compared

to other media genres because it attracts a larger audience and because aspects of

entertainment like humor and fiction disarm peoples’ natural resistance to persuasive

communications about sensitive subjects like race, ethnicity, religion and other group


       In attracting audiences, entertainment has a clear edge over more sober mediums

of influence like lectures or billboards. If people do not pay attention to a message it

cannot affect them; “reception” is the obvious first step of persuasion (McGuire, 1986).

Whether attention to entertaining mediums is more likely to change intergroup prejudice

or behaviors is a topic that has received little rigorous research. One field experiment

showed that in a diversity training, students appreciated games and a film much more

than a lecture and group discussion. However, participants in the lecture and discussion

condition were able to generate more examples of racism by the end of the training

(Sedlacek, Troy, & Chapman, 1976).

       Humor is one significant type of entertainment, but researchers interested in

intergroup prejudice have devoted more research toward understanding how humor

masks, expresses, and perpetuates prejudice and hostility (Dollard, Miller, Doob,

Mowrer, & Sears, 1939; Greenberg & Pyszczynski, 1985; Thomas & Esses 2004;

LaFrance & Woodzicka, 1998) than how humor might improve intergroup relations. For

example, one theory posits that among more prejudiced individuals humor that disparages

stigmatized groups relaxes norms of conduct and thus encourages discriminatory

behavior (Ford & Ferguson, 2004). But on the positive side, prejudiced humor seems to

make low prejudiced people feel guilty and motivated to compensate for their enjoyment

of the joke (Vance, 1998).

       The idea that parodying prejudiced people may be an effective tool was

challenged by a study of the American television show All in the Family. The study found

that low and high prejudiced people interpreted the bigoted character Archie Bunker in a

way that reinforced their prior intergroup attitudes—people with high prejudice saw him

as a humorous role model, and lower prejudiced understood Archie as a negative example

(Vidmar & Rokeach, 1974). I am aware of no research on the power of ridicule—humor

directed against prejudice—which was the type of humor employed by the reconciliation

soap opera’s “town fool” character, who made fun of people seeking to cause trouble

between the two communities. The centrality of humor in much entertainment, and its

established link to prejudiced attitudes and behaviors, indicates a need for future research

on humor as a force for deconstructing intergroup prejudice.

         Like all soap operas and many other media programs, the present intervention

used a fictional storyline. Research supports the assertion that “media create a world

which then becomes our reality” (Harris, 2004, p. 50); people view fictional characters as

real (Esslin, 1982), and fictional information is cognitively processed through the same

channels as nonfiction (Prentice & Gerrig, 1999). Information in fictional depictions can

influence real-world beliefs and attitudes (Prentice & Gerrig, 1999). A working

explanation of this phenomenon is that fiction is viewed less critically compared to

communications understood to be factual. This explanation finds support in the finding

that when fiction overlaps a great deal with an individual’s real social context they

examine it more critically and are less easily persuaded by it (Prentice Gerrig & Balis,


         More research should explore the conditions under which fictional stories can

influence intergroup prejudice, and for how long. A prediction from the literature on

persuasion is that because fictional information is processed peripherally, changed

attitudes may not endure (Petty & Cacioppo, 1986). But other theory and research

suggests that fiction is a compelling method for interventions addressing sensitive topics

like intergroup prejudice. For one, fiction’s ability to create a realistic, if vicarious, sense

of feeling or doing (Gerrig, 1993) means that a fictional story about intergroup friendship

or peacebuilding may approximate the experience of actually having that friendship or

peacebuilding experience. In a series of studies supportive of this idea, researchers read

British schoolchildren stories about fictional British children who befriended refugee and

disabled children. The vicarious experience of that friendship increased children’s

expression of positive attitudes toward these groups (Cameron & Rutland, 2006;

Cameron, Rutland, Brown & Douch, 2006).

       Fictional stories about intergroup prejudice may also facilitate prejudice and

conflict reduction because they offer ways to think about and discuss real life conflict in

less threatening or personal terms. I observed in the present research that the fictional

reconciliation story licensed listeners to express feelings about intergroup relations in

Rwanda that were otherwise difficult to express (given constitutional bans on speech

about ethnicity, the proximity of neighbors they did not trust, etc.) For example, some

study participants stated opinions like “we’ve had leaders here in our village like the

leaders from Muhumoro” (the home of the soap opera’s demagogic villains). This

observation has not been treated to much research, although one study showed that when

Israeli and Palestinian individuals learned about a nonfiction story of a conflict removed

from their own sociopolitical context (Northern Ireland) the vicarious experience of the

other conflict helped participants to let go of some defensiveness and to consider the

other side when discussing their the Israeli-Palestinian conflict (Lustig, 2003). Studying

the impact of fictional narrative communicated through various types of media, and of

entertainment media in general, seems to be a rich area for the study of intergroup

prejudice reduction.

       The social experience of media. The social experience of media consumption is a

universal one—people gather around radios, in front of televisions, in movie theaters.

Being part of a physical audience can increase people’s active participation in the

program, when they are moved to comment to one another, and to discuss the program

during and afterward, as the present research demonstrated. However, the presence of

others may also detract from attention to or enjoyment of the program. Media effects

research and intergroup relations research converge on the point that social groups play a

central role in any persuasive intervention (Katz & Lazarsfeld, 1955; Sherif, 1948). How

much does the social aspect of media consumption amplify or diminish its effects in the

case of changing intergroup relations?

       One way to approach this question is to investigate how more and less social

interaction during media consumption changes the effect of the media intervention. For

example, what if future studies provided extra incentives or opportunities for people to

discuss the program, such as with a quiz show that followed the program? Boosting

discussion of a media program could allow investigators to estimate the “added value” of

discussion for a show that aims to affect intergroup prejudice. Another approach could be

to study media programs that have a more or less social format, such as a talk radio

program vs. a radio documentary that covered the same material about prejudice or

conflict reduction. Talk radio has been dubbed an “electronic backyard fence” that

simulates neighbors talking to one another (Douglas, 2000). Thus, when people listen to

programs that feature public discussions are they more likely to believe that the messages

promoted on that show represent a social consensus? Is the effect of discussion broadcast

through media similar to private discussions of media messages about intergroup

relations? The simplest research design could investigate the difference between social

and solitary media consumption. For example, what is the difference between listening to

the radio alone in the car vs. in the car with someone else?

       Radio: the role of listening and imagination. While research has considered

differences among the influence of print, audio, and film on attitude and behavior change

(e.g., Andreoli & Worchel, 1978), no sizeable body of evidence weighs one way or the

other as to which is more persuasive and under what conditions. Meta-analyses of

persuasion effects reveal no reliable differences among audio, text, and visual modalities,

most likely because the number of experiments conducted with audio or visual modalities

are relatively small compared to those delivered textually (e.g., Kumkale & Albarracıon,

2004; Johnson & Eagly, 1989). Most laboratory experiments use written vignettes as the

experimental manipulation.

       One intriguing quality of radio is that it is a “blind medium” (Douglas, 1999).

Listeners are forced to imagine people and events heard over the radio. Theory describing

the power of mental imagery over human behavior has a distinguished pedigree in

psychology, reaching back to William James, who stated “every [mental] representation

of a movement awakens in some degree the actual movement which is its object” (1890,

p. 526). By imagining actions portrayed on the radio, do listeners stimulate their own

behaviors because of this presumed relationship between the mental representation of

behavior and actual behavior? Contemporary research on mental representations of action

and of even more abstract concepts like goals suggests that the link between our mental

structures and our social behavior is perhaps even stronger than previously postulated

(see Bargh, Chen & Burrows, 1996; de Vignemont & Singer, 2006). It would be a

productive theoretical and also practical exercise to document the effects of a subtle

intervention involving radio and mental imagery, contrasted with the effects of overt

interventions to shape intergroup relations, including education and legislation.

       While the lack of visual images in the blind medium of radio leaves a great deal

of room for listeners’ imagination, auditory cues can also activate stereotypes that guide

listeners to fill in the blanks with positive or negative inferences. Cantril and Allport

(1936) found that people use stereotypes to visualize people speaking on the radio. One

counterintuitive prediction to test with future research is that stereotyping speakers on the

radio may have the ironic effect of improving the listener’s attitudes toward the speaker’s

group. Take the example of a listener who hears a radio interview with a person

introduced as an illegal immigrant. The label “illegal immigrant” will most likely activate

the listener’s stereotypes of illegal immigrants, and she may visualize an exemplar of an

illegal immigrant who is representative of those stereotypes (Smith & Zarate, 1992) One

effect of this visualization is that the listener may react negatively to the speaker, if the

stereotypes are negative (e.g., she visualizes an uneducated, unsanitary person who is

taking away jobs from her fellow nationals.) At the same time, if the interview with the

illegal immigrant disconfirms the listener’s mental image (e.g., she is struck by the

speaker’s intelligence, decorum, or distinguished personal history) the listener’s attitudes

toward illegal immigrants in general may shift. This prediction is based on the idea that

group stereotypes and attitudes rest partly on people’s memory of salient exemplars, or

representative members of the group. If impressions of these exemplars change, it is

possible that impressions of the entire group can change as well (Rothbart & John, 1985;

Sherman, 196, Smith, 1990). Exploring the power of listening and generating mental

imagery for changing intergroup attitudes or inspiring positive intergroup behaviors

would be an interesting path for future research.

        Field research promotes a broader conception of intergroup relations. Intergroup

prejudice and conflict in Rwanda involves a host of interrelated phenomena like mistrust,

ethnic stereotypes, historical narratives about ethnic origins, intergroup violence, and

trauma. All over the world, problematic intergroup relations are characterized by

different combinations of these and other factors. However contemporary psychological

research focuses on milder forms of ingroup preference and favorable discrimination

extended to the ingroup, to the exclusion of more hostile forms of prejudice that involve

factors like violence, trauma, and historical narrative (Brewer, 1999; cf. Green et al.,

1999; Lee, McCauley, Moghaddam, & Worchel, 2004). As this study demonstrated, field

research is well situated to study these interrelated facets of intergroup prejudice and

conflict. In addition, field research also opens up questions about larger social forces that

are difficult to study in the laboratory, like social networks and shifting political fortunes.

       A useful direction for future research would be to directly measure a population’s

rates of trauma and aggression after an intergroup relations intervention like the present

one had targeted intergroup norms, attitudes, and related behaviors in that society. In

addition, psychologists could benefit from investigating historical narratives about the

origins of group identity or hierarchy that explain and justify reasons for intergroup

prejudice and violence (e.g., Kiernan, 1996; Lustig, 2003). Investigating whether

alternative narratives about groups could shift general ideologies about the fairness and

natural order of intergroup hierarchy could contribute to existing research on the etiology

of intergroup attitudes (e.g., Jost & Banaji, 1994; Sidanius & Pratto, 1999).

Limitations of current research

       Given the distinctiveness of this study’s context and methods, two questions of

external validity are in order. First, would norms shift only within our listening groups

and not in larger communities? That is, did the experimental listener groups create

conditions in which norms could shift more easily, as opposed to existing communities

that are less compact and communal in their radio listening?

       Experimental listening groups were meant to be recreations of the small groups of

neighbors that listen together within larger communities, not recreations of the

communities themselves. Moreover, real world small listener groups are probably more

rather than less likely to support the norm shifts observed in this experiment (Fishkin,

1995). The size (40 people) and heterogeneity (due to random selection) of the artificially

created listener groups should diminish listeners’ propensity to discuss the program and

to express their opinions, compared to smaller real world groups of sympathetic

neighbors or family that listeners normally listen with.

       One could question whether changes in perceived norms within small listener

groups can spread to the larger community and society. Through interviews with other

members of the communities where we conducted our research, we discovered that

research participants did discuss the radio program with other members of the community

outside of the experiment (Paluck, 2006). This extra-group discussion suggests a

possible social influence model in which the norm perceptions change in small dedicated

listener groups, which then influence other members of their village, thus spreading local

norms out to the community or regional level. Future research should map these influence


       Although these data are from Rwanda, there is nothing uniquely Rwandan about

the pattern of social norm perception and norm-consistent behavior they reveal. However,

given the media’s special significance in Rwanda, it is reasonable to pose a second

question about external validity: could a mass communication program be as influential

in countries without the same history of media influence, or with more media outlets?

Certainly, very few cases parallel Rwanda’s hate radio during the genocide, in which

“…the radio became for most people the sole source of news as well as the sole authority

for interpreting its meaning” (Des Forges, 1999, p. 70). However, radio looms less large

in Rwanda’s post conflict environment, where the media’s legacy may bias Rwandans

toward greater suspicion of the media, and in which there are more radio stations to

choose among.

       Still, what of environments that are more media saturated? Hundreds of media

outlets are not necessarily indicative of a diversity of factual and normative information.

Take the United States, where mainstream television and radio media share many

characteristics that suggest a predominance of certain norms and values, not to mention

stylistic and presentational formulas. Despite the hundreds of media programs available

in the U.S., the fact of their relative homogeneity makes it reasonable to question (as

some have, Anderson & Bushman, 2002; Lithwick, 2006) whether mainstream television

programming and news shows are affecting social norms such as tolerance for violence

and political torture in American society. Moreover, while the present study was

conducted at a national level, it is truly a study of small groups of listeners and their local

norms, which can be found in every society where small groups of friends or family

choose similar entertainment and news outlets. In the end, whether media effects on

prejudice translate to more developed, Western societies is an empirical question, one

that so far has not been explored in any depth or with much rig9or. But translation to

Western countries is by no means the yardstick for measuring the utility of these

findings—many dozens of countries in the world are currently home to a range of media

outlets comparable to that of Rwanda.


       The present research provides some of the first clear evidence of the media’s

impact on intergroup prejudice and conflict. It is, moreover, one of the first large scale

demonstrations that a field experiment can be deployed to estimate the causal impact of a

media program on intergroup relations. The results of a yearlong radio program aimed at

reconciliation in Rwanda reveal the utility of psychological perspectives on social norms,

conformity, and peer influence for understanding media influence on intergroup prejudice

and conflict. The radio program influenced social norms governing positive intergroup

behavior but failed to change personal beliefs—perhaps the same process by which media

may escalate prejudice and conflict. Future research should continue to measure the

media-prejudice-conflict relationship in real world settings, while grappling with the

basic social psychological mechanisms that seem to drive its effects.


Abdalla, LL.B., & Torrey, N. (1999). Conflict Mapping and Media Programs
       Assessment: The Case of Liberia’s Talking Drum Studio. Common Ground

Abdalla, Shepler, S. Hussein, S.(2002). Evaluation of Talking Drum Studio- Sierra
       Leone. Common Ground Productions.

Aboud, F. E., & Doyle, A. B. (1996). Does talk of race foster prejudice or tolerance in
      children? Canadian Journal of Behavioral Science, 28, 161-170.

Adorno,T.W., Frenkel-Brunswik, E., Levinson, D.J. & Sanford, R.N. (1950) The
      authoritarian personality. N.Y.: Harper.

Allport, G. (1954). The Nature of Prejudice. Reading, MA: Addison-Wesley.

Andersen, S.M., Downey, G., Tyler, T. (2005). Becoming engaged in community:
      Personal relationships foster social identity. In G. Downey, C. Dweck, J. Eccles,
      & C. Chatman, Social identity, coping, and life tasks (pp. 210-251). New York:
      Russell Sage Foundation.

Anderson, C.A., & Bushman, B.J. (2002). The effects of media violence on society.
      Science, 295, 2377-2378.

Andreoli, V., & Worchel, S. (1978). Effects of media, communicator, and message
      position on attitude change. Public Opinion Quarterly, 42, 59-70.

Arceneaux, K. (2005). Using cluster randomized field experiments to study voting
      behavior. Annals of the American Academy of Political and Social Science, 601,

Asch, S. (1958). Effects of Group Pressure upon the Modification and Distortion of
       Judgments. In E.E. Macoby et al. (Eds.) Readings in Social Psychology, (3rd ed.,
       pp.174-183). New York: Holt, Rinehart, and Winston.

Article 19, (1996). Broadcasting Genocide: Censorship, Propaganda & State-Sponsored
        Violence in Rwanda 1990-1994. London: Article 19.

Ball-Rokeach, S., Rokeach, M., & Grube, J.W. (1984). The Great American Values Test:
       Influencing Behavior and Belief Through Television. New York: The Free Press.

Banaji, M. R. (2001a). Ordinary prejudice. Psychological Science Agenda, American
        Psychological Association, 14 (Jan-Feb), 8-11.

Banaji, M. R. (2001b). Implicit attitudes can be measured. (In H. L. Roedeger, III, J. S.
        Nairne, I. Neath, & A. Surprenant (Eds.), The nature of remembering: Essays in
        honor of Robert G. Crowder. (pp.117-150). Washington, DC: American
        Psychological Association.)

Banaji, M. R., & Hardin, C. D. (1996). Automatic stereotyping. Psychological Science.
        Vol 7(3), 136-141.

Bandura, A. (2006, September). Going Global with Social Cognitive Theory. Cherlin
      Memorial Lecture, Yale University.

Bandura, A. (2004). Social cognitive theory for social and personal change enabled by
      the media. In A. Singhal, M. Cody, E. Rogers, & M. Sabido, (Eds.),
      Entertainment-Education and Social Change. Mahwah, NJ: Lawrence Erlbaum.

Bandura, A. (2002). Environmental sustainability by sociocognitive deceleration of
      population growth. In P. Schmuch, & W. Schultz (Eds.). The Psychology of
      Sustainable Development. (pp. 209-238). Dordrecht, the Netherlands: Kluwer.

Bandura, A. (1986). Social Foundations of Thought and Actions: A social cognitive
      theory. Englewood Cliffs, NJ: Prentice-Hall.

Banks, J.A., & Banks, C.M. (2003), Handbook of Research on Multicultural Education.

Bargh, J.A., Chen, M. & Burrows, L. (1996). The automaticity of social behavior: Direct
       effects of trait concept and stereotype activation on action. Journal of Personality
       and Social Psychology, 71, 230-244.

Batson, C. D., Fultz, J., & Schoenrade, P. A. (1987). Distress and empathy: Two
       qualitatively distinct vicarious emotions with different motivational
       consequences. Journal of Personality, 55(1), 19-39.

Batson, C., Polycarpou, M.P., Harmon-Jones, E., Imhoff, H.J., Mitchener, E.C., Bednar,
       L.L., Klein, T.R., & Highberger, L. (1997). Empathy and attitudes: Can feeling
       for a member of a stigmatized group improve feelings toward the group? Journal
       of Personality and Social Psychology, 72, 105-118.

Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in
      Experimental Social Psychology, Vol. 6, 1-62. New York: Academic Press.

Blanchard, F. A., Crandall, C. S., Brigham, J. C., & Vaughn, L. A. (1994). Condemning
      and condoning racism: A social context approach to interracial settings. Journal
      of Applied Psychology, 79, 993-997.

Blanchard, F. A., Lilly, T., & Vaughn, L. A. (1991). Reducing the expression of racial
      prejudice. Psychological Science, 2, 101-105.

Blumer, H. (1958). Racial Prejudice as a Sense of Group Position. Pacific
      Sociological Review 1, 3-7.

Blumer, J.K., & Katz, E. (Eds.) (1974). The Uses of Mass Communications: Current
      Perspectives on Gratifications Research. Beverly Hills, CA: Sage.

Bodenhausen, G. V., Kramer, G. P., & Süsser, K. (1994). Happiness and stereotypic
      thinking in social judgment. Journal of Personality and Social Psychology, 66,

Bourgault, L.M. (1995). Mass Media in Sub-Saharan Africa. Bloomington: Indiana
      University Press.

Brenick, A., Lee-Kim, J., Killen, M., Fox, N.A., Leavitt, L., A., Raviv, A. (in press).
       Social judgments in Israeli and Arabic children: Findings from media-based
       intervention projects. In D. Lemish & M. Gotz (Eds.) Children, Media and War.
       Cresskill, NJ: Hampton Press.

Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate?
      Journal of Social Issues, 55 (3), 429-444.

Brusset, E., & Otto, R. (2004). Evaluation of Nashe Maalo: Design, Implementation, and
       Outcomes Social Transformation Through the Media. Channel Research on behalf
       of Search For Common Ground.

Bryant, J., & Thompson, S. (2002). Fundamentals of Media Effects. Boston: McGraw

Bushman, B. J., & Anderson, C. A. (2001). Media violence and the American public:
     Scientific facts versus media misinformation. American Psychologist, 56, 477-

Cameron, L., & Rutland, A. (2006). Extended contact through story reading in school:
      Reducing children's prejudice toward the disabled. Journal of Social Issues, 62(3),

Cameron, L., Rutland, A., Brown, R., & Douch, R. (2006). Changing children's
      intergroup attitudes toward refugees: Testing different models of extended
      contact. Child Development, 77(5),1208-1219.

Campbell, D. T. (1969). Reforms as experiments. American Psychologist 24, 409-429.

Cantril, H, & Allport, G. (1935). The Psychology of Radio. New York: Harper &


Carey, J.W. The Chicago School and mass communication research. In E.E. Dennis & E.
       Wartella (Eds.), American Communication Research: The Remembered History.
       Mahwah, N.J.: Erlbaum, pp 21-38.

Chalk, F. (1999). Hate radio in Rwanda. In H. Adelman & A. Suhrke, (Eds.), The Path of
       A Genocide: The Rwanda Crisis from Uganda to Zaire. New Brunswick, NJ:
       Transaction Press.

Chein, I. (1946). Some considerations in combating intergroup prejudice. Journal of
       Educational Sociology, 19(7), 412-419

Chirot, D., & Seligman, M.E.P. (Eds.) (2001). Ethnopolitical Warfare: Causes,
        Consequences, and Possible Solutions. Washington, D.C.: APA Press.

Chrétien, J., Dupaquier, J., Kabanda, M., Ngarambe, J., & Reporters Sans Frontières.
       (1995). Rwanda: Les médias du génocide. Paris: Karthala.

Crandall, C.S., Eshleman, A., & O’Brien, L.T. (2002). Social norms and the expression
      and suppression of prejudice. Journal of Personality and Social Psychology, 82,

Crandall, C.S., & Stangor, C. (2005). Conformity and Prejudice. In Dovidio, J.F., Glick,
      P., & Rudman, L.A. (Eds.), On the Nature of Prejudice: Fifty years after Allport.
      Malden, MA: Blackwell.

Cooper, E., & Jahoda, M. (1947). The evasion of propaganda: How prejudiced people
      respond to anti-prejudice propaganda. Journal of Psychology, 23, 15-25.

Costrich, N., Feinstein, J., Kidder, L., Maracek, J., & Pascale, L. (1975). When
       stereotypes hurt: Three studies of penalties for sex-role reversals. Journal of
       Experimental Social Psychology, 11, 520-530.

Des Forges, A. (1999). Leave None to Tell the Story: Genocide in Rwanda. New York:
      Human Rights Watch.

Davies, P.G., Spencer, S.J., & Steele, C.M. (2005). Clearing the air: Identity safety
       moderates the effects of stereotype threat on women's leadership aspirations.
       Journal of Personality and Social Psychology, 88(2), 276-287.

Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled
      components. Journal of Personality and Social Psychology. 56, 5-18.

Devine, P.G., & Elliott, A.J., (2000). Are racial stereotypes really fading? The Princeton

       trilogy revisited. In Stangor, C., (Ed.), Stereotypes and Prejudice: Essential
       Readings. Philadelphia: Taylor & Francis.

Dollard J., Doob, L. W., Miller, N. E., Mowrer, O. H., & Sears, R. R. (1939). Frustration
       and Aggression. New Haven: Yale University Press.

Donovan, R.J., & Leivers, S. (1993, Summer). Using paid advertising to modify racial
      stereotype beliefs. The Public Opinion Quarterly, 57(2), 205-218.

Doob, L. W. (1935). Propaganda: Its psychology and technique. New York: Henry Holt.

Douglas, S.J. (2000). Listening In: Radio and the American Imagination. Minneapolis:
      University of Minnesota Press.

Dovidio, J. F., Kawakami, K., Johnson, C., Johnson, B., & Howard, A. (1997). On the
      nature of prejudice: Automatic and controlled processes. Journal of Experimental
      Social Psychology, 33, 510-540.

Gaertner, S. L., Dovidio, J. F., Anastasio, P. A., Bachman, B. A., & Rust, M. C. (1993).
       The common ingroup identity model: Recategorization and the reduction of
       intergroup bias. In W. Stroebe & M. Hewstone (Eds.), European review of social
       psychology (Vol. 4, pp. 1-26). Chichester, UK: Wiley.

Environics Research Group. (2001). Citizenship Advertising Campaign “Canada:
       We All Belong”, (PN5048, contract 6C501-010874/001/CB.) Communication

European Centre for Conflict Prevention, (2003). The Power of the Media. Utrecht:

Feldman,S. (1988). Structure and consistency in public opinion: The role of core beliefs
      and values. American Journal of Political Science, 32(2), 416-440.

Feshback, N.D. (1988). Television and the development of empathy. In S. Oskamp,
      (Ed.) Television as a Social Issue. Newbury Park: Sage Publications.

Festinger, L.A. (1957): A Theory of Cognitive Dissonance. Stanford, CA: Stanford
       University Press.

Fishkin, J. S. (1995). The Voice of the People. New Haven, CT: Yale University Press.

Fiske, S.E. (2005). Social cognition and the normality of prejudgement. In Dovidio, J. F.,
        Glick, P., & Rudman, L. A. (Eds.) On the nature of prejudice: Fifty years after
        Allport. Malden, MA: Blackwell Publishing.

Flowerman, S.H (1949). The use of propaganda to reduce prejudice: a refutation.

       International Journal of Opinion and Attitude Research, 3, 99-108.

Frazier, M. (2006, March 9). Lights camera learn: City of Decatur uses movie Crash as
        part of diversity training. Illinois Herald & Review.

Frohardt, M., & Temin, J. (2007). The use and abuse of media in vulnerable societies. In
       Thompson, A., The Media and the Rwandan Genocide. Pluto Press/Fountain
       Publishers/ IDRC.

Gaertner, S. L., & Dovidio, J. F. (2000). Reducing intergroup bias: The Common Ingroup
       Identity Model. Philadelphia, PA: Psychology Press.

Gatwa, T. (1995). Ethnic conflict and the media: The case of Rwanda. Media
       Development, 3, 18–20.

Gardiner, G. S. (1972). Complexity training and prejudice reduction. Journal of Applied
       Social Psychology, 2, 326-342.

Gerrig, R.J. (1993). Experiencing Narrative Worlds. New Haven: Yale University Press.

Goffman, E. (1959). The Presentation of Self in Everyday Life. Anchor Books.

Gorn, J.G, Goldberg, M.E., & Kanungo, R.N. (1976). The role of educational television
       in changing the intergroup attitudes of children. Child Development, 47, 227-280.

Gourevitch, P. (1998). We Wish to Inform You that Tomorrow We Will be Killed with
      our Families. Farrar, Straus, & Giroux.

Green, D. P., Abelson, R.P., & Garnett M.. 1999. The Distinctive Political Views of
       Hate-Crime Perpetrators and White Supremacists. In Deborah A. Prentice and
       Dale T. Miller (eds.), Cultural Divides: Understanding and Overcoming Group
       Conflict. New York: Russell Sage Foundation, pp. 429-64.

Gurin, P., Peng, T., Lopez, G. E., & Nagda, B. A. (1999). Context, identity and
       intergroup relations. In D. Prentice & D. Miller (Eds.) Cultural divides:
       Understanding and overcoming group conflict (pp. 133-170). New York: Russell
       Sage Foundation.

Graves, S.D. (1999). Different and the Same: A final report. Unpublished report,
       Carnegie Corporation, US.

Greenberg, J., & Pyszczynski, T., (1985). The effect of an overheard ethnic slur on
      evaluations of the target: how to spread a social disease. Journal of Experimental
      Social Psychology, 1, 61-72.

Hagengimana, A., Hinton D., Bird, B., Pollack, M., Pitman, R.K. (2003). Somatic panic-

       attack equivalents in a community sample of Rwandan widows who survived the
       1994 genocide Psychiatry Research 117, 1–9.

Hagengimana, A., Hinton, D. ().Ihahamuka, a Rwandan Syndrome of Response to the
      Genocide: Blocked Flow, Spirit Assault, and Shortness of Breath.

Hegarty, P., & Pratto, F. (2001). The effects of category norms and stereotypes on
       explanations for inter-group differences. Journal of Personality and Social
       Psychology, 80, 723–735.

Hendy, D. (2000) Radio in the Global Age. Cambridge, UK: Polity Press.

Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup Bias. Annual Review of
      Psychology, 53(1), 575-604.

Horovitz, B. (1993, January 19). Harmonic convergence, racial tolerance is suddenly a
       hot topic in advertising. Los Angeles Times, The Times Mirror Company.

Hovland, C.I., Janis, I.R., Kelley, H.H. (1953). Communication and Persuasion:
`     Psychological studies of opinion change. New Haven and London: Yale
      University Press.

Hovland, C. I., Lumsdaine, A., & Sheffield, F. (1949). Experiments on Mass
      Communication. Princeton: Princeton University Press.

Hovland, C. I., & Sears, R.R. (1940). Minor studies of aggression: correlations of
      economic indices with lynchings. Journal of Psychology 9, 301-10., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007., retrieved February 8, 2007.
                                                                                         109, , retrieved February 8, 2007., retrieved January 30,

Hur, K.K., & Robinson, J.P. (1978)The social impact of Roots. Journalism Quarterly, 55,

Iyengar, S., & Kinder, D.R. (1987). News That Matters: Television and American
       Opinion. University of Chicago Press.

James, W. (1890). Principles of Psychology. New York: Holt.

Janis, I.L., Kaye, D., Kirschner, P. (1965). Facilitating effects of “eating-while-reading”
         on responsiveness to persuasive communications. Journal of Personality and
         Social Psychology, 1, 181-186.

Jennings-Walstedt, J., Geis, F.L., Brown, V. Influence of television commercials on
       women's self-confidence and independent judgment. Journal of Personality and
       Social Psychology, 38(2), 203-210.

Johnson, B.T., & Eagly, A.H. (1989). Effects of involvement on persuasion: A meta-
      analysis. Psychological Bulletin,106(2), 290-314

Jost, J., & Banaji M.R., (1999). The role of stereotyping in system-justification and the
         production of false consciousness. British Journal of Social Psychology, 33(1), 1-

Katz, I., & Hass, R. G. (1988). Racial ambivalence and American value conflict:
        correlational and priming studies of dual cognitive structures. Journal of
        Personality and Social Psychology, 55, 893-905

Katz, E., & Lazarsfeld, P. (1955). Personal Influence: The part played by people in the
       flow of mass communication.. New York: The Free Press.

Katz, P.A., & Zalk, S.R. (1978). Modification of children’s racial attitudes.
       Developmental Psychology, 14(5), 447-461.

Katz, P. (2000). Research Summary. Intergroup Relations Among Youth: Summary of a
       Research Workshop. Carnegie Corporation, NY.

Kellow, C.L., & Steeves, H.L. (1998). The role of radio in the Rwandan genocide.
      Journal of Communication, 107-128).

Kelman, H. (1997). Social-psychological dimensions of international conflict. In W.

       Zartman and L. Rasmussen (Eds.), Peacemaking in International Conflict;
       Methods and Techniques. Washington: United States Institute of Peace.

Kinder, D. (1998). Communication and opinion. Annual Review of Political Science, 1,

Kinder, D. R., & Sears, D. O. (1981). Prejudice and politics: Symbolic racism versus
       racial threats to the good life. , 414-431.

Kuran, T. (1995). Private Truths, Public Lies: The Social Consequences of Preference
       Falsification. Cambridge: Harvard University Press.

LaBenevolencija (2004). Design document, radio serial drama on the roots of
      mass violence, genocide, trauma and trauma healing and reconciliation.
      Unpublished working manuscript. Kigali and Amsterdam: LaBenevolencija.

LaFrance, M., & Woodzicka, J.A. (1998). No laughing matter: Women's verbal and
      nonverbal reactions to sexist humor. In Swim, Janet K, & Charles Stangor, (Eds.)
      Prejudice: The Target's Perspective. San Diego: Academic Press. (pp. 61-80).

Lasswell, H.D (1928), The function of the propagandist, International Journal of Ethics,
      38(3), 258-68.

Latané, B. (1996). Dynamic Social impact: The creation of culture by communication.
       Journal of Communication, 46, 13-25.

Lazarsfeld, P.F. (1947). Some remarks on the role of mass media in so-called tolerance
       propaganda. Journal of Social Issues, 3(3), 17-25.

Lee, Y., McCauley, C., Moghaddam, F., Worchel, S. (Eds.) (2004). The Psychology of
       Ethnic and Cultural Conflict. Westport, CT: Praeger Publishers/Greenwood
       Publishing Group.

Lemarchand, R. (1970). Rwanda and Burundi. London: Pall Mall.

Levy, S., West, T. L., Ramirez, L. F., Pachankis, J. E. (2004). Racial and ethnic prejudice
       among children. In J. Chin, (Ed), The psychology of prejudice and discrimination:
       Racism in America (Vol. 1, pp. 37-60). Westport, CT: Praeger Publishers.

Lewicki, R. J., & Wiethoff, C. (2000). Trust, trust development, and trust repair. In M.
      Deutsch & P. Coleman (Eds.), The Handbook of Conflict Resolution: Theory and
      Practice. San Francisco, CA: Jossey-Bass. 86-107.

Lewin, K. (1952) Group decision and social change. In G. E. Swanson, T. M. Newcomb,
       & E. L. Hartley (Eds.) Readings in Social Psychology.

Leyens, J-P., Herman, G., & Dunand, M. (1982). The influence of an audience upon the
      reactions to filmed violence. European Journal of Social Psychology, 12(2), 131-

Liebkind, K., & McAlister, A. L. (1999). Extended contact through peer modelling to
       promote tolerance in Finland. European Journal of Social Psychology, 29(5-6),

Lippmann, W. (1922). Public Opinion. New York: The Free Press.

Lithwick, D. (2006, September). Photo finish: How the Abu Ghraib photos morphed from
       scandal to law. Slate. Retrieved November 3, 2006, from

Lopez, G. E., Gurin, P. & Nagda, B. A. (1998). Education and understanding structural
       causes for group inequalities. Journal of Political Psychology, 19(2), 305-329.

Malhotra, D., & Liyanage, S. (2005). Long-Term effects of peace workshops
      in protracted conflicts. Journal of Conflict Resolution, 49(6), 908-924.

Maoz, I. and D. Bar-On (2002). From working through the holocaust to current ethnic
       conflicts: Evaluation of the TRT group workshop in Hamburg. Group. 26, 29-48.

Mays, L., Henderson, E.H., Seidman, S.K., Steiner, V.J. (1975). An evaluation report on
      Vegetable Soup: The effects of a multi-ethnic children's television series on
      intergroup attitudes of children. Unpublished manuscript, Department of
      Education, State of New York, Albany.

Mays, V.M., Bullock, M., Rosenzweig, M.R., Wessells, M. (1998). Ethnic conflict global
      challenges and psychological perspectives. American Psychologist, 53(7), 737-

McCann, I.L., & Pearlman, L.A. (1990). Psychological trauma and the adult survivor:
     Theory, therapy, and transformation. New York: Brunner/Mazel.

McCaul, K.D., Jacobson, K., Martinson, B. (1998). The effects of a state-wide media
     campaign on mammography screening Journal of Applied Social Psychology,.
     28(6), p. 504.

McGuire, W. J. (1985). Attitudes and attitude change. In G. Lindzey & E. Aronson
      (Eds.), Handbook of. Social Psychology. New York: Random House.

McGuire, W.G. (1986). The myth of massive media impact: Savagings and salvagings. In
      G. Comstock (Ed.), Public Communication and Behavior. Orlando, FL: Academic

Melvern, L. (2004). Conspiracy to Murder: The Rwanda Genocide. Verso Hardback.

Mendelberg, T, (2002). The deliberative citizen: Theory and evidence. In M.X. Delli
      Carpini, L. Huddy, and R.Y. Shapiro (Eds.), Political Decision Making,
      Deliberation and Participation: Research in Micropolitics, 6. Greenwich, CT:
      JAI Press.

Milgram, S. (1974). Obedience to Authority. Harper & Row.

Miller, G. (1969). Psychology as a means of promoting human welfare. American
        Psychologist, 24, 1063–1075.

Miller, D.T., Monin, B., & Prentice, D.A., (2000). Pluralistic ignorance and inconsistency
        between private attitudes and public behaviors. In D.J Terry & M.A. Hogg (Eds.),
        Attitudes, Behavior, and Social Context: The role of norms and group
        membership. Mahwah, NJ: Erlbaum.

Mitnick, L.L., & McGinnies, E. (1958). Influencing ethnocentrism in small discussion
       groups through a film discussion. Journal of Social and Abnormal Social
       Psychology. 56, 82.

Mutz, D., (1998) Impersonal Influence: How Perceptions of Mass Collectives Affect
       Political Attitudes. Cambridge: Cambridge University Press.

Mutz, D. (2002). Cross-cutting social networks: Testing democratic theory in practice.
       American Political Science Review, 96(2), 111-26.

Mutz, D., & Martin, P.M. (2001). Facilitating communication across lines of political
       difference: The role of mass media. American Political Science Review, 95(1), 97-

Myers, D.G., & Bishop, D.G. (1970). Discussion effects on racial attitudes. Science,
       169(21), 778-779.

Newbury, C. (1988). The Cohesion of Oppression: Clientship and Ethnicity in Rwanda,
     1860-1960. New York: Columbia University Press.

Noelle-Neumann, E. (1973). Return to the concept of powerful mass media. Studies of
       Broadcasting, 9, 67-112.

Nzyuko, S. (1996). Does research have any role in information/education/communication
      programs in Africa? An insiders’s view. Journal of Health Communication, 1(2),
Oskamp, S. (1989). Nuclear war on television: The impact of 'The Day After.'" In G.
      Comstock (Ed.), Public Communication and Behavior. Orlando, FL: Academic

Overdulve, C.M. (1997). Fonction de la langue et de la communication au Rwanda. Neue
      Zeitschrift für Missionswissenschaft, 53(4), 271-283.

Paluck, E.L. (2006). The second year of a “new dawn”: Year Two evidence for the
       impact of the Rwandan reconciliation radio drama Musekeweya. Internal report
       produced for LaBenevolencija, Kigali / Amsterdam.

Paluck, E.L., & Green, D.P. (2007, accepted for publication) Prejudice reduction: What
       do we know? A critical look at evidence from the field and the laboratory. Annual
       Review of Psychology.

Paluck, E.L., Green, D.P., & Nzamukwereka, A (2004). Evaluating LaBenevolencija’s
       “Reconciliation Radio” Campaign: Analysis and Report of Evaluation Study
       Pre-tests. Report for LaBenevolencija Rwanda Office. Yale University.

Pearlman, L.A. (2001). The treatment of persons with complex PTSD and other trauma-
      related disruptions of the self. In: Wilson J.P., Friedman M.J., and Lindy, J.D.,
      (Eds.), Treating Psychological Trauma & PTSD. New York: Guilford Press, pp.

Pedersen, A., Walker, I., & Wise, M. (2005). ‘‘Talk does not cook rice’’: Beyond anti-
       racism rhetoric to strategies for social action. Australian Psychologist, 40(1), 20–

Prentice, D., Gerrig, R., & Bailis, D.S. (1997). What readers bring to the processing of
       fictional texts. Psychonomic Bulletin and Review, 4, 416-420.

PRI: Penal Reform International (2004). The Righteous: between oblivion and
       reconciliation? Example of the province of Kibuye. Report on Monitoring and
       Research on the Gacaca. Paris / Rwanda: PRI.

R. C., & Thurstone, L.L., (1933). Motion Picture and the Social Attitudes of
        Children. New York: Macmillan.

Petty, R.E., & Cacioppo, J. T. (1986). Communication and persuasion: Central and
        peripheral routes to attitude change. New York: Springer/Verlag.

Prentice, D.A., & Gerrig, R.J. (1999). Exploring the boundary between fiction and
       reality. In S. Chaiken & Y. Trope, (Eds.), Dual-Process Theories in Social
       Psychology. New York: Guilford.

Prentice, D.A., Gerrig, R.J., & Bailis, D.S. (1997).What readers bring to the processing of
       fictional texts. Psychonomic Bulletin and Review, 4, 416-420.

Prosecutor v. Nahimana, Barayagwiza, & Ngeze, ICTR-99-52-T (2003).

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

Richeson, J., & Trawalter, S. (2005). On the categorization of admired and disliked
       exemplars of admired and disliked racial groups. Journal of Personality and
       Social Psychology. 89(4), 517–530

Rogers, Everett. (1995). Diffusion of Innovations. Fourth edition. New York, NY: The
       Free Press.

Rogers, E.M., Vaughan, P.W., Swalehe, R.M. A., Rao, N., Svenkerud, P., & Sood, S.
       (1999). Effects of an entertainment-education radio soap opera on family planning
       in Tanzania. Studies in Family Planning, 30(3), 193-211.

Rojas, H., Shah, D.V., Cho, J., Schmierbach, M., Keum, H., Gil-De-Zuniga, H. (2005).
       Media dialogue: Perceiving and addressing community problems. Mass
       Communications & Society, 8(2), 93-110.

Rokeach, M. (1971). Long-range experimental modification of values, attitudes, and
      behavior. American Psychologist, 26, 453-459.

Rokeach M. (1968). Beliefs, Attitudes, and Values. San Francisco: Jossey-Bass, Inc.

Rokeach M. (1973). The Nature of Human Values. New York: Free Press.

Rosin, H. (2006, June 5). Life Lessons. New Yorker Magazine.

Rothbart, M. & John, O.P. (1985). Social categorization and behavioral episodes: A
      cognitive analysis of the effects of intergroup contact. Journal of Social Issues,
      41(3), 81-104.

Rubin, D.B. (1990). Formal modes of statistical inference for causal effects. Journal of
       Statistical Planning and Inference. 25,279-292.

Rubin, A.M., & Perse, E.M. (1987). Audience activity and soap opera involvement.
       Human Communication Research, 14, 246-268.

Rudmin, L.A. 2004: Sources of implicit attitudes. Current Directions in Psychological
      Science, 13(2), 79-82.

Ruscher, J. B. (2001). Prejudiced communication: A social psychological perspective.
      New York, NY: Guilford Publications.

Salovey, P., & Williams-Piehota, P. (2004). Field experiments in social psychology:

       Message framing and the promotion of health protective behaviors. American
       Behavioral Scientist, 47, 488-505.

Schaller, M., Asp, C. H., Rosell, M. C., & Heim, S. J. (1996). Training in statistical
       reasoning inhibits the formation of erroneous group stereotypes. Personality and
       Social Psychology Bulletin, 22, 829-844.

Schecter, H., & Salomon, G. (2005). Does vicarious experience of suffering affect
       empathy for an adversary? The effects of Israelis’ visits to Auschwitz on their
       empathy for Palestinians. Journal of Peace Education, 2(2), 125–138.

Scheepers, E., Christophides, N.J., Goldstein, S, Usdin, S., Patel, D.S., & Japhet, G.
      (2004). Health communication—a holistic overview of the impact of Soul City.
      Health Promotion Journal of Australia, 121-129.

Schofield, J.W, & Pavelchak, M.A. (1989). Fallout from The Day After: The impact of a
       TV film on attitudes related to nuclear war. Journal of Applied Social Psychology,
       19(5), 433-448.

Sears, D.O. (1986) College sophomores in the laboratory: Influences of a narrow data-
       base on social psychology's view of human nature. Journal of Personality &
       Social Psychology 51, 515-530.

Sears D.O., Whitney R.E. (1973). Political persuasion. In Handbook of Communication,
       I Pool & W Schramm (Eds.), pp. 253–63. Chicago: Rand McNally.

Sedlacek, W.E., Troy, W., & Chapman, T. (1976). An evaluation of three methods of
       racism-sexism training. Personnel and Guidance Journal, 196-198.

Shapiro, M.A., & Chock, T.M. (2003). Psychological processes in perceiving reality.
       Media Psychology, 5, 163-198.

Sherif, M. (1936). The Psychology of Social Norms. New York: Harper & Brothers.

Sherif, M., & Hovland, C.I. (1961). Social Judgment: Assimilation and Contrast Effects
        in Communication and Attitude Change. New Haven: Yale University Press.

Sherif, M. & Sherif, C.W. (1953). Groups in harmony and tension: an integration of
        studies on intergroup relations. New York: Harper.

Sherman, J.W. (1996). Development and mental representation of stereotypes. Journal of
      Personality and Social Psychology, 70(6), 1126-1141.

Sidanius, J., & Pratto, F. (1999). Social Dominance: An Intergroup Theory of Social
       Hierarchy and Oppression. New York: Cambridge University Press.

Singer, J.L., & Singer, D.G. (1976). Fostering creativity in children: Can TV stimulate
       imaginative play? Journal of Communication, 26, 74-80.

Singhal, A., & Rogers, E.M., (2004). The status of entertainment education worldwide. In
       Arvind Singhal, Cody, M.J., Rogers, E.M., & Sabido, M. (Eds.) (2004).
       Entertainment-education and social change: History, research, and practice.
       Mahwah, NJ: Lawrence Erlbaum Associates, p. 3-18.

Singhal, A., Cody, M.J., Rogers, E.M., & Sabido, M. (Eds.) (2004). Entertainment-
       education and social change: History, research, and practice. Mahwah, NJ:
       Lawrence Erlbaum Associates.

Smith, E. R. (1990). Content and process specificity in the effects of prior experiences. In
       T. K. Srull & R. S. Wyer, Jr. (Eds.), Advances in Social Cognition (Vol. 3, pp. 1–
               59). Hillsdale, NJ: Erlbaum.

Smith, E. R., & Zarate, M. A. (1992). Exemplar-based model of social judgment.
       Psychological Review, 99, 3–21.

Sood, S., Menard, T., & Witte, K. (2004). The theory behind entertainment-education. In
       A. Singhal, M. Cody, E. Rogers, & M. Sabido, (Eds.), Entertainment-Education
       and Social Change. Mahwah, NJ: Lawrence Erlbaum.

Stangor, C., & Sechrist, G. B., and Jost, J. T. (2001). Social Influence and intergroup
       beliefs. In J. P. Forgas and K. D. Williams (Eds.), Social Influence. (pp. 235-252).
       Philadelphia: Psychology Press.

Stephan, W. G., & Finlay, K. (1999). The role of empathy in improving intergroup
       relations. Journal of Social Issues, 55(4), 729-744.

Stephan, W. G. & Stephan, C. (1984). The role of ignorance in intergroup relations. In N.
       Miller & M. Brewer (Eds.) Groups in Contact: The Psychology of Desegregation.
       New York: Academic Press.

Stephan, W.G., & Stephan, C.W. (2001). Improving Intergroup Relations. Thousand
       Oaks, CA: Sage Publications.

Staub, E. (1989). The Roots of Evil: The Origins of Genocide and Other Group Violence.
       Cambridge: Cambridge University Press.

Staub, E., Pearlman, L.A., Weiss, G. (in preparation). Public Education through Radio to
       Prevent Violence, Promote Trauma Healing and Reconciliation, and Build Peace:
       Rwanda and the Great Lakes Region of Africa. Working draft, University of

Straus, S. (under review). What Is the Relationship between “Hate Radio” and

       Violence? Rethinking Rwanda’s “Radio Machete.” University of Wisconsin,
       Madison. Paper available from author.

Straus, S. (2006). The Order of Genocide: Race, Power, and War in Rwanda. Cornell
        University Press.

Tajfel, H., Turner, J. C. (1979). An integrative theory of intergroup conflict. In S.
        Worchel, W. G. Austin (Eds), The Social Psychology of Intergroup Relations.
        Monterey, Brooks/Cole.

Temin, J. (2001). A review of peace building initiatives through media in Africa.
       Unpublished manuscript, US Agency for International Development.

Thomas, C.A., & Esses, V.M. (2004). Individual differences in reactions to sexist
     humour. Group Processes and Intergroup Relations, 7, 89-100.

Thompson, A., (2007). The Media and the Rwandan Genocide. Pluto Press/Fountain
     Publishers/ IDRC.

United Nations Development Program (2004). Human Development Report 2004:
       Cultural Liberty in Today’s World. New York: UNDP.

Vance, K. M. (1998). How Self-Affirmations Influence the Motivation and Behaviors
       Associated with Prejudice Reduction. Dissertation Abstracts International: Section
       B: The Sciences and Engineering. Vol 59(5-B), pp. 2490.

Verwimp, P. (2004). Death and survival during the 1994 genocide in Rwanda. Population
      Studies, 58(2), 233-245.

de Vignemont, F., Singer, T. (2006). The empathic brain: how, when and why? Trends in
       Cognitive Science, 10(10) 435-441.

Vrij, A., Akehurst, L., Smith, B. (2003). Reducing ethnic prejudice: An evaluation of
        seven recommended principles for incorporation in public campaigns. Journal of
        Community & Applied Social Psychology, 13, 284–299

Vrij, A., & Smith, B. (1999). Reducing ethnic prejudice by public campaigns: An
        evaluation of a present and new campaign. Journal of Community & Applied
        Social Psychology, 9, 195–215.

Vrij, A., Van Schie, E., & Cherryman, J. (1996). Reducing ethnic prejudice through
        public communication programs: A social-psychological perspective. The Journal
        of Psychology, 130, 413–420.

Warren, T.C. (2005). Rebellious beliefs: Communication network structure and the

       mobilization of ethnic insurgency. Paper presented at the Laboratory in
       Comparative Ethnic Processes (LiCEP) special meeting, San Diego, CA,
       December 9-10, 2005.

Wood, W. (2000). Attitude change: Persuasion and social influence. Annual Review of
      Psychology, 51, 539–570.

Zillman, D. (2006). Dramaturgy for emotions from fictional narration. In J. Bryant & P.
       Vorderer (Eds.), Psychology of Entertainment. Mahwah, NJ: Lawrence Erlbaum

Zillman, D., & Bryant, J. (1983). Uses and effects of humor in educational ventures.
       In P. E. McGhee &J. H. Goldstein (Eds.), Handbook of Humor Research: Vol. 2.
       Applied Studies. New York: Springer-Verlag, pp 173-193.

Zimbardo, P.G. (2004). Does psychology make a significant difference in our lives?
      American Psychologist, 59(5,) 339–351.

                                       Appendix A

                                     Field visit form

Site ville de ………………………………………………….
Date de la visite…………………………………..
Chercheur ………………………………………………….
Durée de la visite: de ……………… à ……………….. TOTAL: …1h…. 40 Min
Episodes …… à .……. de MUSEKEWEYA / URUNANA
nombre des personnes présentes: ……
“Réactions Notables”:
    •   Rire (de quoi):
    •   Surprise (de quoi):
    •   Désapprobation (sur quoi):
    •   autres commentaires (et comment les participants réagissent sur des commentaires
        de leur collègues).
Des personnages ou des histoires qu’ils semblent aimer:
Des personnages ou des histoires qu’ils semblent détester:
D'Autres Observations :
Commentaires et suggestions qu'ils font à vous le chercheur :

Tous problèmes ou événements peu communs pendant la visite.

Toute autre chose que vous voulez rapporter ?

                       pas du tout   un peu   beaucoup

    beaucoup moins un peu moins la même chose un peu plus beaucoup plus

                                                         Comparé à          Comparé à
                                     Le groupe          l’autre groupe   d’autres sites dans la
comment intéressé
comment enthousiaste
comment distrait
comment confus
Combien ils l’ont discute


N B. Ecrivez tous les commentaires que les participants font durant la période de
rafraîchissement concernant le programme : combien de temps parlent-il du
programme ? Durant le rafraîchissement : 5%, 15%, 25%, 35%, 50%. 75%, 100% ou
autre pourcentage que vous estimez.

                                                       Appendix B
Statistical Model

        Three nested statistical models are used to estimate the reconciliation program
effect. Due to the random assignment of the program, all three models provide unbiased
estimates; they differ in terms of the precision with which the treatment effect is
estimated. The relative precision of the models is not known ex ante, however. This
section therefore describes the potential advantages and disadvantages of each approach.
In the next section, which presents the empirical results, we show that the choice among
models is substantively inconsequential, as all three produce very similar estimates and
significance levels.

       The most elementary of the three models is a bivariate regression of the
dependent variable on a dummy variable for treatment, scored 1 if individuals were
exposed to the reconciliation program and 0 if they were exposed to the health program.
The dependent variable in this context could be any of several outcome measures that
were gauged through surveys, focus groups, or observation of group deliberations.
Denote the dependent variable for individual i as Yi and the treatment as Xi. The average
treatment effect of the reconciliation program is represented by the parameter β.
Unobserved causes of the dependent variable are denoted ui. The bivariate model takes
the familiar form

Yi = " + !X i + ui .                                                                 (eq.1)

        The unbiasedness of the least squares estimate of β depends on two assumptions.
The first is that Xi and ui are statistically independent, an assumption satisfied by random
assignment of the clusters containing the Xi. The second assumption is that the Xi are
measured without error. This assumption is violated to the extent that respondents in the
reconciliation group did not in fact listen to the program; it is violated also to the extent
that respondents in the health group listened to the reconciliation program. I have reason
to believe that the Xi contain relatively little measurement error. Attendance of the
reconciliation group at the monthly listening visits was strong throughout, and spot
checks of the health group revealed no evidence that the respondents were familiar with
content of the reconciliation show. It should be noted that to the extent that this type of
mismeasurement did occur, our treatment effects will underestimate the absolute value of

        The second model builds on the bivariate model by adding dummy variables for
the blocks within which randomization occurred. The strata dummy variable for block j
and individual i is denoted Sij. Thus, the regression model for the six strata is

Yi = # + "X i + ! 1 S i1 + ! 2 S i 2 + ! 3 S i 3 + ! 41 S i 4 + ! 5 S i 5 + ui .     (eq.2)

        Inclusion of these strata dummies is essential when the units of assignment are
assigned to the reconciliation group with probabilities that vary across strata. That was
not the case here. Across all of the strata (Kibuye, Gisenyi, Kigali Ngali, Butare, Batwa,

and Survivor villages), the probability of assignment into the treatment group was 50%.
The strata dummies are therefore optional. In the event that they lower the root mean
squared error of the regression, their inclusion will lower the standard errors associated
with the estimated treatment effect. On the other hand, if their predictive power is low,
the standard errors associated with the estimated reconciliation program effect may grow
larger due to the loss of degrees of freedom.

       The model may be further expanded to included pre-treatment covariates (Zi1
…ZiK) that were measured in the baseline survey.

Yi = $ + #X i + " 1 S i1 + " 2 S i 2 + " 3 S i 3 + " 41 S i 4 + " 5 S i 5 + ! i1 Z i1 + ... + ! iK Z iK + ui   (eq.3)

        These covariates are statistically independent of the (random) program
assignment. In finite samples, however, these covariates may be correlated with the
program assignment due to sampling variability. Excluding these covariates does not
lead to bias, since random assignment ensures unbiased estimation of the reconciliation
program effect, but controlling for covariates may help improve the precision with which
the program’s effect is estimated. Again, improvement in precision depends on whether
the gains in predictive accuracy obtained by including the covariates offsets the
corresponding loss of degrees of freedom.

        When estimating the standard errors of the program effect, each of these models
must make special allowance for the fact that random assignment was conducted at the
village level, not the individual level. This adjustment is implemented using STATA’s
“robust cluster” option, which takes account of the fact that individuals were assigned
within village-level clusters. It turns out, however, that this adjustment generates
standard errors that scarcely differ from conventional standard errors because there is
very little evidence of intra-cluster correlation among individuals. The regression
estimates presented below include both conventional and robust-cluster standard errors
for each of the three types of models. The point of including all six permutations of
models and standard errors is to emphasize the robustness of our results. In the end, it
makes very little difference how the reconciliation program effect and its standard error
are estimated.

        One final consideration arises due to the fact that my dependent variables are
survey responses that are expressed in a series of ordered categories. Applying linear
regression to these data imposes arbitrary assumptions about the cardinal values attached
to each response option. A standard remedy for this problem is to use ordered probit
(Greene 2002). I present ordered probit regressions based on the three specifications
presented in this section, but it should be noted that the results are not materially different
from those obtained using linear regression.

Qualitative analyses To analyze the qualitative data, I translated (from French to
English) and typed transcripts of all focus groups into a database, along with identifying
codes for site, speaker, composition of the group (e.g. male, female, average age), and
experimental condition. Prior to the study, I (with the LaBenevolencija staff) enumerated

hypothetical, desired answers and anticipated common answers to focus group questions
that might be expected with and without the influence of the radio program. These
prototypical answers became a priori coding categories for the open-ended data.

        Once the database was compiled, I read through the open-ended responses (blind
to experimental condition) to see how well the a priori codes fit the dataset. Some
responses did not fit into the a priori coding schema or were numerous enough to merit
their own code. For example, one proverb about intermarriage was mentioned so
frequently it was awarded its own code. The new codes added made the final coding
scheme inductive as well as a priori—focused on the treatment hypotheses but also fully
descriptive of the participants’ discussions.

With this coding system, I coded each spoken turn in every focus group. Each turn could
receive from 0 to n codes, n being the total number of codes pertaining to the comment.
For certain questions, the first round of coding revealed that the coding system was too
fine-grained to classify large enough groups of responses for statistical analysis. I
recombined the codes according to their theme (e.g., in response to the question about
intermarriage, “individual benefits” was a compound code from “it changes the parents of
the couple” and “it changes the couple’s attitudes.”). Finally, I submitted the codes to
probit regressions to analyze for treatment effects using the same model laid out above.

                                        Appendix C

Individual questionnaire items and frequency distribution

Belief system
Evolution of violence: “Mass violence comes out of small actions, like spreading
       negative ideas about a group of people, or stealing from them.”
                74% strongly agree
                19% somewhat agree
                4% somewhat disagree
                4% strongly disagree
Onset of violence: “Violence like the violence that happened in Rwanda in 1994 comes
       about suddenly.”
                48% strongly agree
                13% somewhat agree
                9% somewhat disagree
                29% strongly disagree
Scapegoating: “People who cannot meet their physical and psychological needs are more
       likely to blame someone else for their problems.”
                31% strongly agree
                36% somewhat agree
                16% somewhat disagree
                18% strongly disagree
Active bystandership: “If I stand by while others commit evil actions, I am also
                56% strongly agree
                19% somewhat agree
                9% somewhat disagree
                16% strongly disagree
Intermarriage: “When people marry each other from different (regions,
       religions, ethnicities, etc) this contributes to the peace”
                72% strongly agree

              20% somewhat agree
              5% somewhat disagree
              3% strongly disagree
Trauma: “Traumatized people are crazy”
              13% strongly agree
              16% somewhat agree
              19% somewhat disagree
              52% strongly disagree
Trauma: “Perpetrators of violence can also be traumatized by their own actions.”
              59% strongly agree
              34% somewhat agree
              3% somewhat disagree
              4% strongly disagree
Trauma healing: “Traumatized people can recover”
              54% strongly agree
              35% somewhat agree
              8% somewhat disagree
              4% strongly disagree
Trauma healing: “Recovering from grief (intimba) and from trauma (ihungabana) may
       take a very long time.”
              64% strongly agree
              29% somewhat agree
              5% somewhat disagree
              3% strongly disagree
Health: “A pregnant woman who has AIDS can be given a chance to have
       a healthy baby.”
              Yes: 80%
Health: “You can share something with someone who has AIDS.”
              Yes: 92%

Descriptive and prescriptive norms of radio programs
Intermarriage in family: ‘I tell my children (or I will tell my future children) they
       should only marry people from the same group (regions, religions, ethnicities) as
                  25% strongly agree
                  17% somewhat agree
                  11% somewhat disagree
                  47% strongly disagree
“It is naïve to trust people”
                  13% strongly agree
                  16% somewhat agree
                  19% somewhat disagree
                  51% strongly disagree
“There is mistrust”
                  42% strongly agree
                  34% somewhat agree
                  9% somewhat disagree
                  14% strongly disagree
Dissent: “If I disagree with something that someone is doing or saying, I should keep
                  27% strongly agree
                  16% somewhat agree
                  12% somewhat disagree
                  45% strongly disagree
Trauma healing: “It is better for my mental health to never talk about the experiences that
       have caused me great pain and suffering”
                  29% strongly agree
                  20% somewhat agree
                  12% somewhat disagree
                  39% strongly disagree
Tests for pregnant women: “It’s necessary that every woman who is pregnant goes to the

       health center to be tested”
               Yes: 78%
Behavioral intentions
Social distance: “Think about a person from a different religion,
       region, or ethnicity who has done harm to a person from your group in the past.
       Would you be willing to, (1) greet, (2), work, (3) share a beer with, (4), allow
       your child to marry a child of, this person?” (combined index of four levels)
               39% strongly agree
               39% somewhat agree
               12% somewhat disagree
               9% strongly disagree
Empathy index: “Do you ever try to imagine the thoughts or feelings of other people who
       you don’t know in Rwanda” (prisoners, survivors, the poor, leaders).
               35% strongly agree
               57% somewhat agree
               7% somewhat disagree
               1% strongly disagree

Frequency distributions include both reconciliation and health program conditions


  With the exception of psychologists interested in the influence of violent media on
aggression (e.g., Bushman & Anderson, 2001) and in health messages communicated
through the media (e.g., McCaul, Jacobson, & Martinson, 1998; Salovey & Williams-
Piehota, 2004).
  Currently, psychological research on prejudice focuses on inter- or intra-individual
levels of explanation of phenomena such as identity, values and stereotypes, while
recognizing the mass media as a component of ambient societal prejudice that individuals
absorb (e.g., Devine, 2000;; Rudmin, 2004) or a
carrier of prejudiced or stereotypic messages that may trigger such intra-individual
processes, (e.g., Davies, Spencer, & Steele, 2005).
  This line is sometimes blurry; for example, one must decide whether gathering
passersby on the street into a parked truck to judge television commercials (Vrij et al.,
1996) is a field study (I decided it was not), and whether gathering children into a
classroom to watch a cartoon together several times over many months (Mays et al.,
1979) is a laboratory study (I decided it was not.) The general rule I use for these “grey
areas” is to judge whether the circumstances of the media consumption simulate to a
reasonable degree the natural condition of media consumption, and whether the
participants in the study are the people who would actually consume the media program.
In the case of the classroom study, it is reasonable to think that children would watch a
cartoon series at their school; it is not likely at all that most of us will watch television
advertisements in a stranger’s van parked on the side of the road, even though that road is
not in a laboratory.
  The usual criticisms of laboratory research as an analogue to the world also apply: the
measurement of behaviors is difficult, laboratories usually rely on special populations
(Sears, 1986), participants are highly aware of the measurement, there is usually no
follow up to test for longitudinal effects (see also Paluck & Green, 2007).
  The translation of Western psychological theory into Rwandan culture is a constant
challenge, even in a literal sense (all shows are written and broadcasted in Kinyarwanda,
Rwanda’s common language). Although space does not permit a detailed consideration of
the application of Western psychology in sub-Saharan Africa, my research with this
specific program is inevitably part of a larger debate about the authenticity and validity of
such collaborations. This evaluation was not designed to test this specific question,
though the author’s personal opinion is that this radio program is not a case of unilateral
imposition of Western theory upon local understandings.
 This logic was also used to predict changes resulting from World War II films shown to
American soldiers; the prediction was that soldiers would understand the material and

that motivations and behaviors would follow this understanding (Hovland, Lumsdaine, &
Sheffield, 1949).
  The message harkens to the contact hypothesis in psychology, but does not incorporate
all of its conditions of equal status, authority sanction.
  I selected two communities from each (pre-2006) province on the basis of a close match
on literacy rates, gender and religion ratios, radio ownership, and dwelling type (e.g.
isolated vs. communal living) according to the 2002 Rwandan census. Geography was
also factored into the decision— communities had to be accessible by at least a dirt road,
and the matched communities could not be neighbors (to prevent the groups from
influencing one another). I drew general population communities from the following four
key provinces in Rwanda. The first is northwest Gisenyi, the seat of Hutu extremism
leading up to the genocide and at the time a relatively less stable zone of the country due
to incursions by ex-genocidaires living across the border in the Democratic Republic of
Congo. The second is western Kibuye, home to massacres of an exceptional scale due to
large concentration of Tutsis prior to 1994. The third is southern Butare, home of the
national university and the site of most resistance to the genocide, which eventually gave
way to some of the largest massacres. The fourth is northern Kigali Ngali, where RPF
soldiers invaded, ending the genocide but precipitating reprisal violence.
  Data collection in the two prisons that took part in this study was not approved by the
academic human subjects committee; these data are currently embargoed under ethics
regulations and will not be presented here.
     Adult age was defined by official US research guidelines as above the age of 18.
   In some communities, the male-female balance was adjusted to reflect the actual ratio
in the community. For example, survivor communities are up to 80% female.
 From this point forward, “we” refers to myself and the Rwandan research team with
whom I worked in the field.
  58% reported that they listen every day, 11% reported that they listen once per week,
and 24% reported that their listening was sporadic, usually due to the fact that they could
not always afford batteries, or because they listened to others’ radios and were dependent
upon overhearing the radio in community settings or others’ homes.
   A multivariate test involving a regression of treatment on background characteristics
(including dummies for missing data in the pre-test) was, as expected, insignificant,
meaning that there was no reliable association between characteristics of the research
sites and their assignment to the reconciliation or health radio program condition.

   I ruled out the idea of distributing radios to participants because of the likelihood that
many participants would sell the radio and because there was no guarantee they would
listen to the radio program to which they were assigned.
     Local brews made of bananas and fruit.
   Field reports from research assistants indicate that the monthly visits did achieve a
certain amount of similitude with the participants’ everyday experience. After the first
few visits when researchers reported that individuals asked whether there would be a quiz
at the end of the listening session, and that some of the more educated asked if they could
take notes, the participants settled into a routine. The research assistants reported that
they felt comfortable and accepted by the groups. From month to month, the assistants
reported the kind of variation in individuals’ and groups’ mood and reactions that we
would expect if the experience had become unremarkable. For example, on certain
occasions participants were tired, restless, or more interested than on others. After the
first two visits, there is a notable shift in which researchers describe participants relaxing,
talking more, and greeting people who pass by like they would in an ordinary setting in
which they were sitting together listening to the radio. Non-participating individuals from
the community stopped asking to join and did not watch the group from a distance, as
they did on the first visits. There were no differences in these observations between the
health and reconciliation sites.
   The radio NGO was not able to delay broadcast of the program to allow for pre-
broadcasting experimentation, and the reconciliation program was broadcast over the
state radio station, which has nationwide coverage.
   It is possible that other community individuals at the health sites listened to the
reconciliation program and talked about it with our health participants. This possibility
also causes me to underestimate the treatment effect size. It is not ideal for the purposes
of isolating health listeners from the reconciliation program that the program was
available to members of their community over national radio. However, it is an advantage
that reconciliation listeners were able to talk about the program with other members of
their community, because this made the research listening experience more realistic.
  Participants asked our research team to spend more time in their community over the
course of a few days rather than return to the community for a few hours over the course
of many days.
     We also collected other outcome measures, not reported here.
  Government reaction toward researchers conducting a political behavior survey in the
month prior to our data collection forced us to be even more cautious in the way we
broached sensitive subjects in our own study; for example, unlike previous investigations,
the NGO with which we worked decided it was not wise to ask participants to discuss
their own ethnicity.

  Recall that ethnicity was referred to in coded language: for clarity and simplicity I
write “ethnicity” where this happened in the interview questions.
  Due to political sensitivities, it was impossible to ask about ethnic groups, although the
groups “prisoners” and “genocide survivors” roughly map onto those two groups. While
the category “poor people” applies to the majority of the population, is of course relative
and this question measured the ability to look beyond one’s own situation of poverty to
one’s neighbors’ situation.
   Single sex focus groups are often used when there is reason to believe that women
won’t speak as much in the presence of men. However in some communities it was
necessary to mix men and women in focus groups because fewer men were available on
that day and time to participate in the focus group. This happened at the two survivor
sites where the ratio of men to women was particularly low.
   The order of the administration of individual surveys and focus groups was
counterbalanced at each site, such that half of the participants engaged in individual
questionnaires before participating in the focus group and vice versa, to test for order
effects based on which activity was conducted first. For example, one might say that
participants who learned what their neighbors thought in a focus group might respond
differently on a subsequent individual questionnaire, or that participants would
participate in a focus group differently after solidifying their private opinions in an
individual interview exercise. I tested for order effects in our data, and they were not
significant, thus all data were collapsed for subsequent analyses.
   Regarding the sources of mistrust and future trust: Reconciliation focused on the
community as they discussed sources of current mistrust and of future rebuilt trust, while
health groups avoided the topic of their neighbors sitting with them by discussing familial
and broad political sources of mistrust. In response to the question “why is there
mistrust,” reconciliation groups emphasized problems of community reconciliation,
specifically neighbors who do not ask for pardon, who request but do not receive
forgiveness, and insincere apologies (17% of comments in the reconciliation group vs.
2% in the health group.) For example, a survivor in a reconciliation listening group

      The mistrust comes from a lack of reconciliation. We did our best to forgive the
      génocidaire, we even marry their daughters, but some of them still don't ask for
      forgiveness and confess their sin. –(FG interview, #14-19)

      Reconciliation groups also mentioned problems with community dialogue and
discourse: kurenzaho (keeping secrets), hurtful and untrue rumors, and a general lack of
community conversation. The concern with muted conflict, expressed in 17% of the
reconciliation groups’ comments, was not statistically different from the health group
discussions (8%), but is consistent with the previous result that reconciliation participants

were more likely to speak out about the delicate subject of community problems:

      There is mistrust in our community because there are the people who don't want to
      discuss the truth about what happened in this country --(FG interview, #7-5)

      Moreover, reconciliation focus groups reasoned that in the future trust would be
rebuilt from within their community. To a significantly greater degree, reconciliation
groups proposed solutions like community interaction (socializing, sharing resources, and
general engagement between former enemies; 39% vs. 13% of responses) and public acts
of reconciliation (mutual forgiveness, asking pardon, and truth telling about the past; 48%
vs. 25% of responses). Health groups were more likely to say that trust would be
reinforced through government policies: “sensitization” (a common Rwandan term for
governmental campaigns), and laws prohibiting “divisionism” or political favoritism
(28% vs. 4% of responses).
  This particular expression of solidarity raises a cautionary flag for the interpretation of
these results with respect to external validity. If listening groups’ willingness to cooperate
was encouraged by the fact that they had been doing something particular and special—
“coming together all this time” in groups that did not occur naturally—then
generalizations that naturally occurring groups will experience the same degree of
willingness are probably overestimated.
   In all, 1 group decided to keep the stereo for individual use (not listen together); 3
groups proposed but later rejected the idea of individual usage in favor of communal
sharing, and 8 proposed and agreed to manage the stereo and to listen to it communally.
With a one-tailed Mann-Whitney rank sum test, the difference between the reconciliation
and health groups with respect to these proposed and accepted sharing resolutions
reached a significance level of p = .11. Reconciliation program groups were slightly more
likely to propose communal sharing rather than individual, and to agree on the communal
sharing proposal.
  This effect holds even when the responses of genocide survivors are taken out of the
analysis (for the reason that empathy for survivors of violence is more well-developed
among survivors, but this did not prove to be the case.)
  Note that this proposition regarding social norm change offers a facilitating role to
private opinion and knowledge.

Shared By:
yaohongm yaohongm http://