We analyze the validity of the adiabatic approximation, and in particular the reliability of what has been called the standard criterion for validity of this approximation. Recently, this criterion has been found to be insufficient. We will argue tha
t the criterion is sufficient only when it agrees with the intuitive notion of slowness of evolution of the Hamiltonian. However, it can be insufficient in cases where the Hamiltonian varies rapidly but only by a small amount. We also emphasize the distinction between the adiabatic {em theorem} and the adiabatic {em approximation}, two quite different although closely related ideas.
