An explanation is proposed for a long-standing major discrepancy between few-state close-coupling calculations and experiment for excitation of atoms in collision with 'heavy' targets like Ar, Kr and Xe. The discrepancy is associated with singly inelastic collisions in which the projectile is excited but the target remains in its initial state. The explanation of the difficulty lies with the interaction between the active electron in the projectile and the static field of the target. With increasing target size this interaction becomes very strong, with the result that couplings between all projectile states are important. Hence to restrict these couplings to those between a few eigenstates is a mistake. Calculations using the impulse approximation, which does not restrict the projectile-state space, are shown to be in relatively good agreement with the available experimental data for excitation of H(1s) to H(2s) and H(2p) in collision with Kr and Xe targets. This agreement demonstrates clearly the failure of the few-state close-coupling approximations and justifies the interpretation of this failure.