The validity of the stress biases used in reliability studies of high-k dielectric is discussed by analyzing the stress biases used in previous works. For single layer dielectrics, stress biases near the time zero dielectric breakdown point have been used to reduce the test time. However, stacked dielectrics need a more careful approach to avoid overstress. We show that the majority of earlier work on the reliability of high-k dielectric used high electric field and those results may not be optimal for predicting intrinsic reliability characteristics. A simple guideline to avoid overstress is provided.