Accurately identifying developmental language disorder (DLD) in students who speak a home language other than English has proven to be an enduring challenge. Consequently, students with DLD miss out on critical interventions, and those who are acquiring their two languages in a typical manner are placed in settings that do not meet their educational needs. While many factors underlie the issue, language assessment practices are central to the problem and, therefore, to the solution. Standardized tests remain the preferred method despite repeated recommendations against their use with this population due to bias, lack of validation, and limited availability of instruments in the home language. Language sample analysis (LSA), on the other hand, has often been promoted as best practice, but there is a need for empirically based guidance for selecting and interpreting measures that are accurate indicators of DLD. To address this need, three studies were conducted as part of this dissertation. Study 1 is a systematic review of studies that examined the diagnostic accuracy of language sample measures in English revealed several LSA measures and composites that reach 80% sensitivity and specificity for children ages 3 to 10 and at least one measure for all years except age six that reaches 90% or greater. Studies 2 and 3 explored the diagnostic accuracy of a set of LSA measures in English and Spanish, respectively, for use with Spanish-English bilingual 5- and 6-year-olds when adjusting for participants’ relative exposure to each language. Percent grammatical utterances and errors per C-unit in English each yielded 94% diagnostic accuracy for children with at least 70% English exposure. In Spanish, the best model included errors per C-unit and MLU, but it fell just short of the desired 80% threshold for sensitivity and specificity. Results of the three studies are discussed in regard to their implications for clinical practice, language-specificity of clinical markers of DLD, usage-based theory, and future directions for research.
