Test-based educational accountability is widely used in many countries, but is pervasive in the US. Key features of test-based accountability required by the US No Child Left Behind Act are discussed. Particular attention is given to methodological issues such as the distinction between status and growth approaches, the setting of performance targets, compensatory and multiple-hurdle approaches, and the disaggregation of results. The strengths and limitations of growth and of value-added models are discussed. It is concluded that school accountability results are best thought of as providing potentially valuable descriptive information rather than supporting strong inferences about school quality. (Contains 1 table, 1 figure, and 5 notes.)