Lee Branum-Martin, Julie A Washington, Katherine T Rhodes
The items of the DELV-ST did not function well as unidimensional measures of their intended parts in the current sample. Some items and many responses did not function adequately, suggesting that for a better understanding of dialect, a wider variety of linguistic tasks with diagnostically informative responses may be needed. Relations between the DELV-ST and language or reading are therefore questionable.
PURPOSE: Measuring language proficiency in the presence of nonmainstream dialect variation is a crucial task for clinicians and language researchers. Evaluations of the Diagnostic Evaluation of Language Variation-Screening Test (DELV-ST) have suggested that its items do not conform to the recommended scoring rules. The current article evaluates the items from both parts of the DELV-ST.
METHOD: Participants included 853 African American students in Grades 1-5. We used a priori confirmatory factor models to evaluate the published structure versus linguistically based blocks of items for the two parts of the DELV-ST: Language Variation (15 items) and Diagnostic Risk (17 items). We also evaluated models of full multinomial scoring of all item choices.
RESULTS: Most of the items had reasonable sensitivity (factor loadings), although some were weak. The alternate scorings suggest there is substantial information in the variant responses. The multiple factors had only moderate correlations with each other (r = .28-.57).
CONCLUSIONS: The items of the DELV-ST did not function well as unidimensional measures of their intended parts in the current sample. Some items and many responses did not function adequately, suggesting that for a better understanding of dialect, a wider variety of linguistic tasks with diagnostically informative responses may be needed. Relations between the DELV-ST and language or reading are therefore questionable.
SUPPLEMENTAL MATERIAL: https://doi.org/10.23641/asha.33245943.