This paper examines: (1) the use of questions by children at different levels of proficiency in Spanish and English, and (2) the congruency between the language constructs used to measure language proficiency and the natural language repertoire of children as seen in video-tapes of classroom interaction. A quantitative analysis of the data collected on question repertoire indicates that, in general, questions occur more often in the language in which the children are more proficient. The results show no significant difference in the number of questions asked by each child. Discussing measures of language proficiency, it is concluded that a test that measures more than one aspect of language competence is a better predictor of the speaker's communicative competence than one which is limited to a single aspect of that competence. In addition, the paper concludes that tests currently used to measure language proficiency examine aspects of language use that are irrelevant to children's linguistic performance and fail to take into account most of the richness of the children's language repertoire. New test constructs for measuring language proficiency that are more holistic in nature and that are based on what children actually do with language need to be developed. (JK)