Generalized imitation within three response classes in typically developing infants

Effects of modeling and contingent praise on infant imitation of three different responses was analyzed. Generalization to nonreinforced probe models was assessed both within and across response types. Three 12- to 14-month-old infants and their mothers participated in this study. During baseline the mothers provided models only. During treatment mothers modeled and also praised contingent upon infant matching of the training models. During interspersed probe trials the mothers modeled different responses, which, if matched by the infant, produced no praise. The three responses modeled were motor-with-toy, motor-without-toy, and vocal responses. The dependent measure was the percentage of maternal models that were matched by the infant within 6 s. Nonmatching responses of the same response type were also measured. Results showed a systematic increase in the percentages of training and probe models matched by the three infants following the introduction of the model-and-praise treatment condition. Nonmatching responses did not systematically increase. Thus, imitation generalized within response class, but not across response classes.