The following is an excerpt of the intervention methodology. For more information, please see the full text of the article on the publisher's website or through your institution's library.

Experimental context I use a unique field experimental intervention to test whether peer influence can spread intergroup tolerance throughout a social network in the form of intergroup attitudes and behavior. The intervention aims to decrease intergroup prejudice and harassment among teenagers in U.S. high schools. [...]. The current intervention used representatives of various subgroups throughout a high school to model anti-prejudice and anti-harassment behavior such as verbal condemnation of prejudice and confrontation of harassers. [...]

Peer influence intervention We built the field experiment around the Anti-Defamation League's (ADL) A World of Difference Institute Peer Training program, [...]. The program selects and trains a group of students in the participating school to be “Peer Trainers,” whose mandate is to model anti-prejudice attitudes and behaviors throughout the school. [...].

Materials and method [...]. Ten schools agreed to begin the Peer Training program in their school, [...]. We paired each school with its closest match in the sample, using a range of publically available data: number of students per teacher, percentage of students receiving reduced lunch at the school, and the ethnic and racial composition of the school. Next, we randomly assigned one school in each pair to treatment [...] and the other to control [...]. [...]. Pretest [...]. The pretest served to compare the prospective Peer Trainers in each matched pair of schools, and to identify the social network to be interviewed at each school. We asked Peer Trainers about their motivation to join the program and whether they had participated in other clubs, organizations, or groups like the Peer Training program. [...]. To build the rest of the social network sample, we asked prospective Peer Trainers to provide the names of two students they considered close friends and eight students they considered peers [...]. [...]. We used this sociometric information to build two subgroups for the study's outcome measurement: Peer Trainers' close friends (hereafter “Friends”) and their classmates (hereafter “Peers”).

Outcome measures Outcome measurement occurred after the treatment schools had experienced five months of training and before the control schools started training. We mailed letters to the parents of Peer Trainers, Friends and Peers, which introduced the outcome survey as a university-based telephone interview about adolescent viewpoints. [...]. [...]; the survey measured awareness of and attitudes toward intergroup harassment, attitudes toward outgroups and toward political and systemic discrimination and prejudice, and self, peer-reported, and actual anti-prejudice behavior. [...]. We then tested whether anti-prejudice and anti-harassment attitudes and behavior were more common among treatment Trainers and their Friends and Peers.

Awareness To test whether the intervention raised students' awareness of prejudice and harassment in their school, we asked participants when they last overheard teasing or insults about another student's weight. Attitudes [...], interviewers asked participants an open-ended question regarding whether they thought students should intervene if they overheard a person being teased or insulted about their weight. The interviewer also asked if intervening would be effective. Three attitude items related to content from the Peer Training classes regarding structural issues of prejudice and discrimination. [...]. Participants responded to all attitude statements on a scale from 0 (disagree strongly) to 3 (agree strongly).

Behavior First, we asked students to self-report their behavior. On a four point scale from very often to never, we asked participants how often they talked with their friends about the topics of discrimination, prejudice, and bias. Next we asked: “How comfortable do you feel talking with people about [these issues],” presenting a four point scale from extremely uncomfortable to extremely comfortable. Second, we asked participants to report on the behavior of their peers. We asked participants to name which students out of the entire school population (including themselves) were likely to stand up for someone who was getting teased or verbally abused. [...]. We asked students to give full names of up to four people. Each participant received a point if another participant nominated him or her, such that all participants received a nomination score from zero to n−1, with n the total number of students interviewed at their school (self-nominations did not count toward the nomination score). [...]. Third, to observe behavior directly we asked students whether they would post their full name on either, both, or neither of two student-created Internet petitions.[...].