Abstract We report six experiments comparing errorful and errorless learning in the teaching of new information to neurologically impaired adults with severe memory problems. The first experiment is a group study in which amnesic subjects, young controls, and older controls were required to learn two lists of words under two conditions. One of these required subjects to generate guesses that produced incorrect responses, and the other prevented guessing—permitting only correct responses. Conditions and lists were counterbalanced across subjects. People with amnesia scored significantly higher under the errorless condition. We further explored the principle of errorless learning in five single case studies in which severely memory impaired people were required to learn information analogous to that needed in everyday life. Tasks included learning names of objects and people, learning how to programme an electronic aid, remembering orientation items, and learning new items of general knowledge. In each case...