Generalization is a critical outcome for individuals with autism spectrum disorder (ASD) who display new skills in a limited range of contexts. In the absence of proper planning, generalization may not be observed. The purpose of the current study was to directly compare serial to concurrent multiple exemplar training using total training time per exemplar, mean total training time, and exposures to mastery across three children diagnosed with ASD. Additionally, we assessed the efficiency of presenting secondary targets in the antecedent and consequence portions of learning trials and evaluated generalization to tacts not associated with direct teaching. Results suggested that all training conditions produced acquisition and generalization for trained and untrained exemplars. However, the serial multiple exemplar training condition was more efficient for two participants, whereas the instructive feedback condition was the most efficient for the third. Findings are discussed considering previous studies and areas for future research.