The Language Assessment Battery was administered to third to fifth grade limited English speaking students in New York City. Special techniques addressed two common problems in testing these populations: (1) language tests suitable for limited English speakers change their psychometric characteristics over time as the students' language ability improves; and (2) the absence of a comparison population who do not receive the necessary treatment. A cutting score was determined by the intersection of the target and monolingual curves of scores from an earlier test administration. Students scoring above this point more closely resembled monolingual speakers, while students scoring below this point were more like target students in English abilities. The skills typical of the target range of scoring were compared to those which described the monolingual range of scores; Rasch item calibrations and person abilities were examined. Reading skills (as opposed to listening and writing) best differentiated target and monolingual students. This latent trait model was useful in selecting an appropriate norm. The Chapter I Model A1, using test norms as proxy comparison groups, interprets pretest-posttest differences as growth beyond the improvements made by the norms groups. Suggestions for using this model in limited English speaking programs are made. (GDC)