We examined student-, classroom-, and school-level effects in predicting second-grade Spanish-speaking children's oral reading fluency in Spanish. Teachers in 67 randomly selected urban schools administered the Tejas LEE to 1,537 first- and second-grade students. Oral reading fluency was measured in the passages students read for comprehension. Covariates were mean fluency in Grade 1, variability in fluency in Grade 1, degree of grouping in the school, and the proportion of second-grade students in the classroom and/or the school taking the Tejas LEE. Treatment effects were administration format (paper, desktop, handheld) and type of teacher support (no mentoring, web mentoring, and on-site plus web mentoring). Second-grade teachers positively affected students' reading fluency when (a) they administered the Tejas LEE on paper with the associated paper reports in classrooms of bilingual students, and (b) they either received web mentoring and had relatively homogeneous classrooms or received on-site or no mentoring and had ability-grouped classes. Implications for interpreting assessment results are discussed in the context of the type of support provided to teachers and the grouping of bilingual students by language and/or by ability. (Contains 2 tables and 2 figures.)