The purpose of this study was to validate a procedure for sufficient exemplar training that can potentially lead to generalized responding. Based on this procedure, we taught four children with autism, generalized use of regular past tense verbs. We applied a non-concurrent multiple baseline design across the participants. The dependent variable was the number of verbs that needed to be trained for generalization to untrained verbs to occur. The number of verbs that needed to be trained varied considerably across participants as did the number of trials. We concluded that the procedure was validated and that it was easy to implement and it allows for individualization. We discuss the practical implications of our findings for teaching a wide range of skills using the same type of training procedure and data display strategies we provided in this study.