Competence, Autonomy, and Relatedness: Potential Mediators of the Relationship Between Peer Involvement and Classroom Engagement
Method: We analyzed secondary data from a spring 2011 survey of all 9th grade students in one public high school. Surveys were administered electronically in one to two randomly selected classrooms per teacher. (N= 1,080 observations from 436 students in 148 classrooms taught by 91 teachers.) Although mediation is often tested using structural equation modeling, commercial software cannot appropriately model numerous observations from the same student across multiple classrooms. Therefore, we used three-level, cross-classified regression models to test for mediation following procedures for multilevel models (Mathieu & Taylor, 2007; Zhang et al., 2009). Outcomes included behavioral, emotional, and cognitive engagement, defined respectively as observation-specific: 1) positive conduct, effort, and persistence; 2) positive affect; and 3) deep concentration. Mediators were competence, autonomy, and relatedness, defined respectively as observation-specific: 1) ability, 2) interest in learning because of its importance and intention to learn as much as possible, and 3) sense of belonging. Peer involvement aggregated students' perceptions of classmates’ caring, encouragement, and sense of fictive kinship to the classroom-level. Race/ethnicity, sex, socioeconomic status, home language, prior achievement, and aggregate classroom home language and racial composition served as covariates. Unless noted otherwise, all non-dichotomous student- and classroom-level variables were standardized for ease of interpretation.
Results: Because all of the variation in peer involvement occurred between classrooms, potential mediators needed to vary at the classroom-level (Mathieu & Taylor, 2007). Only autonomy met this requirement; 13.4% of its variation occurred between classrooms. Following Zhang et al.'s (2009) procedures, autonomy was significantly and positively associated with behavioral (Π= .547, p<.001), emotional (Π=.421, p<.001), and cognitive (Π= .529, p<.001) engagement. Peer involvement was significantly and positively associated with autonomy (Π= .243, p<.001), behavioral (Π=.125, p<.01), emotional (Π=.387, p<.001), and cognitive (Π=.252, p<.001) engagement. When group mean centered autonomy and aggregate classroom autonomy were entered into models containing peer involvement, the coefficient on peer involvement dropped and became insignificant (behavioral engagement, Π= -.024, p>.05), or less significant (cognitive engagement, Π=.097, p<.05). Sobel tests confirm that autonomy mediated the relationship between peer involvement and behavioral engagement (Sobel=4.647, p<.001) and partially mediated the relationship between peer involvement and cognitive engagement (Sobel=4.533, p<.001).
Implications: These findings extend the self-system process model to adolescent peer contexts. They suggest that classroom-level social work interventions aimed at developing a family-like atmosphere and enhancing students’ reciprocal demonstrations of caring and encouragement may positively influence students’ emotional and cognitive engagement as well as their interest in and intention to learn. This, in turn, may enhance students’ behavioral and cognitive engagement.