A study investigated the use of manipulatives-based mathematics tasks as an alternative method for assessing young Hispanic students with limited English proficiency. Students (n=45) from kindergarten and grades 2 and 3 were administered 14 manipulatives-based tasks. Each task was retested over a period of 2-3 weeks. Task performance was scored on four scales: conceptual understanding; efficient strategy use; accuracy; and fluency. The study addressed research questions of retest reliability, task difficulty, and relationships among subscales, patterns of change in subscales, relationship to general math skills, relationship to language proficiency, and relationship to gender. Results indicate that the manipulatives-based tasks were challenging for this population, despite use of manipulatives and continuous access to them in the classroom. Students improved significantly on six of the 14 tasks over the retesting period, without targeted instruction. Subskills were tightly interrelated at earlier ages, highly differentiated in grade 3, suggesting that manipulative tasks may function differently according to developmental level. Changes from first to second assessment were even across the four criteria subscores. Little relationship was found between math ability and English or Spanish fluency. Small but consistent gender difference was found. (Contains 75 references.) (MSE)