Background: Gaps between dual language learners (DLLs) and monolingual English-speaking children have been reported at kindergarten entry, and many DLL children are classified as long-term English learners, having less access to the general curriculum in the upper grades (NASEM, 2017). The supposed underperformance of DLLs might be due to the language in which they are assessed based on the assumption that the language systems are overlapping and the dominant one for general language ability should account for all knowledge in the specific domain areas. This is although long-standing evidence had shown that DLLs process language thorough a shared conceptual system (Mancilla-Martinez et al., 2021). Despite the extent of literature regarding assessment selections for DLLs, most of the research has focused on literacy and reading outcomes (e.g., NASEM, 2017), leaving the consideration of other domains, such as mathematics, highly unexplored. Purpose: The purpose of this study is to examine whether Latine DLL's early mathematics performance is impacted by the language in which they were assessed (e.g., assessed in English or Spanish based on the preLAS score). Language screener tests have been used to assess DLLs' English proficiency. Head Start uses the Preschool Language Assessment Survey (preLAS, 2000; Duncan & De Avila, 1998) as the screener test to assess DLLs' English proficiency through receptive and expressive language measures. Performance on the preLAS determines children's direct assessment language path (e.g., primarily assessed in English or Spanish). The hypothesis for our study is not straightforward, given that children who passed the preLAS should apparently have the required proficiency to succeed in assessments in English (Thordardottir et al., 2006). However, before elementary school, DLLs might simultaneously develop their vocabulary in both languages, which argues against using one specific language at this age. Setting: This study uses the Head Start Family and Child Experiences Survey (FACES) from 2014 and 2019. Population: The total sample of FACES 2014 is 2,462 children in 60 Head Start programs; of those, 865 were Hispanic/Latino/a based on the child race/ethnicity report provided by parents. Children were between 3 and 5 years old at the time of data collection (M[subscript months] = 48.11; SD = 7.00), and 49.64% were female. A total of 615 children spoke Spanish at home, which is the sample used in this study. The total sample of FACES 2019 is 2,260 children in 59 Head Start programs; of those, 768 were Hispanic/Latino/a based on the child race/ethnicity report provided by parents. Children were between 3 and 5 years old at the time of data collection (M[subscript months] = 47.39; SD = 6.78), and 49.1% were female. A total of 559 children spoke Spanish at home, which is the sample used in this study. Research Design: We use a regression discontinuity design (RDD). Studying the effects of assessing Latine DLLs in English and/or Spanish is a difficult task since students taking the exam in English might differ across multiple unobserved characteristics from students taking the exam in Spanish. An RDD mitigates the concerns about unobserved characteristics (Cattaneo et al., 2021). The rationale in an RDD is that units are assigned to different groups (i.e., treatment and control group) based on their position on a scale or their ranking on a score. The RDD takes advantage of this since an arbitrary cutoff decides who goes to what group. A key consequence is that at the cutoff, the only thing experimenting with a drastic change is the treatment status but not the students' characteristics, which in this study is children's score on the "pre"LAS test. The outcome on this study is Latine DLLs performance in the Woodcock-Johnson III Applied Problems/Woodcock Muñoz Problemas Aplicados assessment. Analysis: We used the language screening assessment (i.e., preLAS) to compare Latine DLLs just above and just below the "cutpoint" that defines whether the math assessment was administered in English or Spanish. In FACES, children scoring eight or more points (out of 20) in the "pre"LAS took the assessments in English, and children scoring less than 8 points took the assessments in Spanish. We transformed this variable, so the cutoff was equal to zero, and -1 means to be 1 point below the cutoff (administered in Spanish), and 1 means to be 1 point above the cutoff (administered in English). The unit of analysis is Latine DLLs; the "pre"LAS is used to define whether children are in treatment or control group, and the outcome is the result of the math assessment. We use a RDD to estimate the effect of taking the math assessment in English by comparing children who were just below the cutoff in the "pre"LAS (control group: took the assessment in Spanish) and children that were just above the cutoff (treatment group: took the assessment in English). Findings: The results show that taking the mathematics assessment in English reduced the score by 1.53 standard deviation units in FACES 2014. This result replicates for FACES 2019, showing that Latine DLLs who took the mathematics assessment in English performed 0.69 standard deviation units lower than those who took the mathematics assessment in Spanish. Thus, Latine DLLs that did pass the preLAS and took the Woodcock-Johnson III Applied Problem in English performed lower than Latine DLLs that did not pass the preLAS and took the Woodcock Muñoz Problemas Aplicados in Spanish, both in FACES 2014 and 2019. Those results are frightening, showing that Latine DLLs are being restricted by the language assessment path. Conclusions: DLL Spanish-English-children bring immense linguistic and cultural assets to schools and communities (Callahan & Gándara, 2014). Yet, this population is widely considered among the most vulnerable of learners (NASEM, 2017). These vulnerabilities are driven by structural inequalities (Umansky & Porter, 2020), such as curriculum design and reliance on assessments that do not reflect Latine DLLs' cultural and linguistic resources, as it is shown by our results. This study will provide the groundwork for meaningful change in policy and practice.