Hidden Bias: How a Language Screener Can Misrepresent Dual Language Learner Performance in Mathematics

Source abstract

Achievement gaps in school readiness between dual language learners (DLLs) and their English-only peers persist, in part because common assessment practices do not fully capture DLLs’ abilities. Using the Head Start Family and Child Experiences Survey, a nationally representative sample of Head Start children, this study examines whether DLLs’ performance varies by assessment language. Applying a fuzzy regression discontinuity design, we found that DLLs who barely failed the language screening and were assessed in Spanish for mathematics outperformed those who barely passed and were assessed in English. These findings reveal biases in assessment practices that may misrepresent DLLs’ abilities, reinforcing educational inequities. Results underscore the need for linguistically responsive assessments that more accurately reflect DLLs’ mathematical competencies.