of English-language learners (ELLs) and other students from culturally non-mainstream backgrounds are longstanding. This article proposes that new paradigms in the research and practice related to ELL testing are needed to address the complexities of language and culture more effectively. Three main areas are identified as key to this paradigm shift: test review, test development, and treatment of language as a source of measurement error. Research examples are provided that illustrate that the proposed paradigm shift is not only necessary but also possible. The authors propose the combined use of generalizability theory and research designs in which ELLs are given the same items in both English and their native languages—an approach that has the potential to reveal more fine-grained understandings of the interactions among first and second language proficiency, student content knowledge, and the linguistic and content demands of test items.