Abstract
Studies of kindergarten pupils have shown that while low intralist similarity results in more rapid acquisition than high intralist similarity there are also more incorrect identifications and false generalizations on transfer tasks. The inference has been that if pupils were taught highly similar words they would make fewer subsequent errors because they would be less prone to fixate on single features of the words. The focus has, however, been on the extremes: high versus low similarity. Data from the present study, designed to extend the earlier results through procedural changes, suggest that moderate similarity may also result in efficient learning. The main implication has to do with the pragmatic matter of preparing beginning reading materials.