Two major evaluations of bilingual education differed significantly in evaluation design and results. The American Institute for Research (AIR) conducted the national impact study for bilingual education and gathered data from many programs. AIR based its study on comparisons of mathematics and language arts achievement of students in bilingual classrooms (with 74% non-English speaking or bilingual children) and control classrooms (with just 17% bilingual or non-English speaking children). The two groups were matched ethnically but not by language dominance. The design of the evaluation of the Individualized Bilingual Instruction (IBI) Program was based on evidence of how the subjects would have performed without the benefit of the program. These data, built up over years, were accumulated from actual pretests of students who enrolled in the program at different ages, thus providing an automatic match for ethnicity, socio-economic level, age, and language dominance. In general, IBI results indicated significantly higher performance for students in its bilingual program than did AIR for Title VII students as a whole. Appendices include the IBI data base from which the paper's findings were taken and technical information regarding IBI testing, data collection, and analysis procedures. (SB)