The federal No Child Left Behind Act mandates a test-based accountability system involving four major elements: goals, performance measures (tests), targets for performance, and consequences for schools success or failure in meeting the targets. Most large-scale tests assess only a limited number of desired outcomes. However, classroom-based tests can usefully complement large-scale tests as instruments to promote educational change and can contribute to greater inferences about student knowledge and skills. Large-scale testing has been shown to have both beneficial and adverse effects on instructional practice, schools, teachers, and students. The author offers general guidelines for improving large-scale test design, improving accountability system design, and incorporating classroom-based assessment into large-scale testing and accountability systems. She cautions that even the most carefully designed accountability system is bound to produce some unintended and perhaps undesirable consequences. It will be important to continue to gather evidence from large-scale studies and the experience of those affected by test-based accountability to inform efforts to improve the system.