A Longitudinal Model for Interpreting Thirty Years of Bilingual Education Research

This article summarizes the results from the re-analysis of two recent bilingual education studies: Greene’s (1998) meta-analysis and Thomas and Collier’s (1997) longitudinal study. First, the article presents the ten main criteria identified over the past three decades by the research community for conducting a methodologically adequate bilingual education study. Second, based on a power analysis of Greene’s data, it is argued that a major reason why many bilingual education studies over the past thirty years remain uninterpretable is because of the high occurrence of Type II errors. Third, based on an effect size analysis of Thomas and Collier’s data, it is maintained that their longitudinal model best explains thirty years of bilingual education research. When Thomas and Collier’s data are converted into effect sizes, their model becomes a longitudinal meta-analysis, and their findings support the results from Greene’s and Willig’s (1985) meta-analyses. This article concludes by making recommendations for enhancing the evaluation of Title VII programs.