RESUMEN
Although much is known about the sources of teachers' self-efficacy (TSE), less attention has been paid to the social-contextual specificity of TSE and the processes influencing the relevance of TSE information sources. This study investigated both dyad-level relationships and the classroom relational climate as predictors of TSE at the student and classroom level. Additionally, we explored two competing hypotheses-assimilation and contrast-articulating how teachers use information conveyed by classroom relationship experiences as a heuristic to interpret relationship experiences with individual students as a TSE source. Elementary school teachers (N = 86; 72.05% female) completed the Student-Teacher Relationship Scale and Student-Specific Teacher Self-Efficacy Scale for randomly selected children (N = 687, 50.1% girls, Grades 3-6) from their classes. Doubly latent multilevel structural equation modeling was used to test for associations of Closeness and Conflict with TSE at both the student (L1) and classroom level (L2). Contextual effects, corresponding to associations of classroom-level Closeness and Conflict with TSE above the same associations at the student level, were calculated to test assimilation and contrast hypotheses. At L1, results indicated positive associations between Closeness and TSE and negative associations between Conflict and TSE. At L2, only Conflict was negatively associated with TSE. Consistent with the contrast hypothesis, the contextual effect of Closeness, but not Conflict, was negative and significant. Hence, teachers' judgments of a relatively close classroom relational climate may lead them to perceive relational closeness with individual students in a more negative light, resulting in lower levels of TSE toward these students.