Evaluations of bilingual education programs have received criticism from many sources. This criticism has been partially due to difficulties in obtaining hard data in bilingual settings, such as the lack of test instruments for bilingual children, the lack of appropriate reference groups, and the deficiencies of traditional evaluation models. The special problems involved in evaluating bilingual programs are described along with emerging solutions. Major problems have been encountered with the three most common evaluation designs: the control group model, the comparison group, and the norm reference group. Alternative evaluation design models include multi-year studies and designs in which children serve as their own control group. Such time series type studies allow for the consideration of developing English skills among the subjects. In evaluating bilingual programs, it is important to avoid rigid models and standardized tests that are likely to underestimate program effectiveness. References are provided. (RW)