In the midst of keen controversies on the impact of high-stakes testing and test-driven external accountability policy, the more balanced and careful selection, interpretation, and use of scientific research evidence are crucial. This article offers a critical synthesis of cross-state causal-comparative and correlational studies that explored the effects of test-driven external accountability policies on reading and math achievement. A meta-analysis of 76 effect-size estimates from 14 selected studies showed a modestly positive effect on average but no significant effect on the racial achievement gap. Nevertheless, this review calls for further evidence on the policy-outcome linkage, revealing limitations, uncertainties, and inconsistencies in many findings. The author explores variations among the studies in terms of independent and dependent variables, analytical samples and methods, and the reporting of statistical and practical significance. Implications for accountability policy and research under the No Child Left Behind Act are discussed. (Contains 6 notes, 2 tables, and 2 figures.)