124 The Professional Counselor | Volume 14, Issue 2 (Interpret), “I feel personally responsible to intervene and assist in resolving bullying incidents” (Accept), “I have the skills to support a student who is being treated disrespectfully” (Know), and “I would say something to a student who is acting mean or disrespectful to a more vulnerable student” (Intervene). Confirmatory factor analyses support the five-factor structure, and convergent validity analyses using the Defending subscale of the Bullying Participant Behaviors Questionnaire (Summers & Demaray, 2008) has been demonstrated by providing positive correlations ranging from .26 to .35 among middle school students (Jenkins & Nickerson, 2016). Researchers have also demonstrated high internal consistency for the subscales among middle school students, with Cronbach’s alpha coefficients ranging from .77 to .87 for the five subscales (Jenkins & Nickerson, 2016). For the current sample, the scales had acceptable internal consistency with Cronbach’s alphas ranging from .66 to .71. For the Interpret subscale, we deleted one item (i.e., “It is evident to me that someone who is being bullied needs help”) to reach an acceptable level of internal consistency (α = .66) for the scale. Defending Behavior We utilized the 3-item Defender subscale of the Participants Roles Questionnaire (PRQ; Salmivalli et al., 2005) to measure defending behaviors students may use to intervene when witnessing bullying. The subscale includes the following items: “I comfort the victim or encourage him/her to tell the teacher about the bullying,” “I tell the others to stop bullying,” and “I try to make the others stop bullying.” Items are rated on a 3-point Likert scale ranging from 0 (never) to 2 (often). Confirmatory factor analyses support the five-factor structure of the PRQ measure, and construct validity has been demonstrated through significant associations between self-reported roles and sociometric status (e.g., popular, rejected, and average), χ2 = 117.7–141.6, all p values < .001, and peer nominations, χ2 = 57.9–88.2, all p values < .001 (Goossens et al., 2006). Among middle school students, the Defender subscale has good internal reliability ranging from α = .79–.93 (Camodeca & Goossens, 2005; Salmivalli et al., 2005). For the current sample, Cronbach’s alpha was high (α = .80). Bystander Status We assessed bystander status by asking participants, “Have you seen bullying at school in the past month?” with response choices Yes and No. The item was developed by the second author, Diana M. Doumas, to assess whether or not students had the opportunity to respond to a bullying incident. Students who reported Yes were classified as bystanders (i.e., the student witnessed bullying and had the opportunity to respond) and students who reported No were classified as non-bystanders (i.e., students who did not witness bullying and, therefore, did not have the opportunity to respond). The item has face validity and researchers have utilized this item previously to measure bystander status among middle school students (Midgett & Doumas, 2020; Moran et al., 2019). In this study, the 30.4% of students who reported Yes to this item at the follow-up assessment (T2) were classified as bystanders, and the 59.6% of students who reported No were classified as non-bystanders. Data Analyses We conducted all analyses using SPSS version 28.0. We imputed missing data and examined all variables for skew and kurtosis. We used a general linear model (GLM) repeated measures multivariate analyses of covariance (RM-MANCOVA) to examine changes in engagement in the five steps of the Bystander Intervention Model between bystanders and non-bystanders across time for the outcome variables Notice the Event, Interpret the Event as an Emergency, Accept Responsibility, Know How to Act, and Decision to Intervene. The independent variables were Time (baseline [T1]; follow-up [T2]) and Bystander Status (bystander; non-bystander). We also controlled for gender, age, and witnessing bullying at baseline. We conducted post-hoc GLM repeated measures analyses of covariance (RMANCOVAs) for each outcome variable. We plotted simple slopes to examine the direction and degree of the significant interactions testing moderator effects (Aiken & West, 1991). We only interpreted
RkJQdWJsaXNoZXIy NDU5MTM1