This study examined the effect of a serial multiple exemplar training (S-MET) procedure on bidirectional naming (BiN) in four preschool children diagnosed with autism spectrum disorder (ASD). A non-concurrent multiple baseline design was used to evaluate the effects of training listener and speaker behavior for one stimulus at a time until BiN occurred. When BiN occurred, probes were conducted to measure whether generalization occurred across settings and people. Three out of four participants' responding met the mastery criterion for BiN, while the fourth participant improved her performance. The results of this study suggest that S-MET may be a promising intervention and contribute to our knowledge about learning histories required for BiN.