Background: Dual language learners (DLLs) -- young children learning and developing in two or more languages -- comprise a third of the U.S. preschool and early elementary student population (MPI Data Hub, 2020). Therefore, identifying the most promising, empirically informed practices to promote DLLs' development is of critical importance. Currently, the predominant approach to supporting DLLs' classroom-based learning is English immersion, in which DLLs are educated entirely in English with virtually no supports for their home language. This approach not only neglects their potential to develop bilingual skills, but it also ignores the growing evidence base showing that bilingual support -- that is, support in both the child's home language and English -- is critical for DLLs' development across domains, including their English language and math skills (Burchinal et al., 2012; Garcia, 2018; Limlingan et al., 2020, 2021; Partika et al., 2021; Valentino & Reardon, 2015; White et al., 2019). Although promising, the evidence on the effectiveness of bilingual supports for DLLs has a few limitations. First, most studies on this topic are correlational and therefore may suffer from omitted variable bias that could arise if, for instance, children with lower English skills were more likely to receive bilingual supports. Second, nearly all studies that examine DLLs' specific classroom experiences have focused exclusively on the preschool year, limiting our understanding of DLLs' experiences throughout early education. Objective: The proposed study aims to fill these gaps by exploiting repeated measures across three years of schooling (preschool-first grade) to examine within-child changes in classroom experiences and English language and math skills, asking: "What is the within-child effect of classroom bilingual supports -- teacher bilingualism, use of Spanish in the classroom, and provision of classroom materials in Spanish -- in early education on DLLs' English language and math skills?" Based on both prior education (e.g., Valentino & Reardon, 2015) and bilingual development (e.g., National Academies of Sciences, 2017) research, bilingual supports are expected to positively predict DLLs' skills across domains. Setting and Participants: Data were drawn from the Tulsa School Experiences and Early Development (SEED) Study, an ongoing longitudinal study on the effects of preschool education for children from economically-disadvantaged families. The present study focuses on the preschool year (2017-2018), when participants attended Head Start centers or Tulsa Public Schools (TPS) school-based pre-k programs, as well as kindergarten and first grade (2018-2020) when children attended TPS elementary schools. The analytic sample includes 479 Spanish-speaking DLLs (operationalized as being from a household where Spanish is spoken). See Table 1 for more details on the sample. Practice: Bilingual Supports: The key practices of interest include three dimensions of classroom bilingual support: teacher bilingualism, teacher's use of Spanish, and provision of materials in Spanish. Each support was coded as a binary (yes/no) variable for each study year. Teacher bilingualism was coded affirmatively if the teacher spoke Spanish fluently. Use of Spanish in the classroom was coded affirmatively if teachers or any of their teaching staff used Spanish in the classroom. Provision of Spanish language materials was coded affirmatively if the classroom had books, print, or literacy materials available in Spanish. Research Design: In the proposed study, within-child fixed effect models were used to identify within-child changes in classroom experiences to predict within-child changes in skills over time, thereby controlling for all "fixed" (time-invariant) characteristics and circumstances of a child. Children's skills included English literacy (Woodcock et al., 2001), expressive vocabulary, and oral comprehension (Semel et al., 2003), as well as math problem-solving (Woodcock et al., 2001). Although child fixed effects greatly control for potential selection bias, a series of classroom covariates (teacher race/ethnicity, age, and years of experience; observed classroom quality [Pianta et al., 2008]; and the proportion of classroom students who are DLLs) and time-varying child and family covariates (relative amount of English vs. Spanish spoken at home, mother's marital status, log of household income, parent education, child age, and whether the child has an Individualized Education Plan [IEP]) were included in the regression as controls. Data Collection and Analysis: Classroom bilingual supports were reported by teachers during the winter of each study year. Children's skills were directly assessed in the fall and spring of preschool and kindergarten and the fall of first grade by trained bilingual assessors over two testing days in a room in the school that was relatively free of distraction. Data on child and family characteristics were collected via annual parent surveys. Separate OLS regression models with child fixed effects were estimated for each support and each outcome, controlling for all aforementioned covariates. Missing covariate data was multiply imputed to account for item-level non-response. Preliminary Results: Bilingual supports were not associated with any English language outcomes; however, they were associated with math problem-solving skills (Table 2). DLL children scored 0.19 SD higher (SE=0.09, p=0.03) when they had a bilingual teacher compared to when they did not have a bilingual teacher, and 0.14 SD higher (SE=0.06, p=0.03) when they were in a classroom that provided materials in Spanish compared to when their classroom did not provide Spanish materials, net of child, family, and classroom covariates. Conclusions: This study is the first to use repeated measures to identify within-child associations between bilingual supports and developmental outcomes for DLLs, across preschool and early elementary school. Results suggest that, for Spanish-speaking DLLs from economically-disadvantaged families, having bilingual teachers and access to Spanish-language materials may promote greater growth in math problem-solving skills compared to having monolingual teachers and limited access to Spanish-language materials, respectively. Although more research is needed that expands beyond a single school district and includes DLLs who speak languages besides Spanish, results provide promising evidence that home language supports can support DLLs' growth in domains besides the home language itself -- specifically, math problem-solving. Next steps include examining if such supports are promotive of additional developmental domains by testing associations between bilingual supports and children's self-regulatory skills. Findings have important implications for how schools can best support DLLs, a large and growing student population traditionally underserved in U.S. public schools.