Imitation and emulation are both important response modalities when learning new tasks. The current study tested the effects of establishing generalized imitation (GI) across missing topographies (gross motor, fine motor, multiple-step motor) on number of sessions-to-criterion for four preschoolers with developmental delays who were learning novel dressing skills. We also recorded participants' response modalities (imitative or emulative) during instruction. We used a pre- and postintervention design with a nested multiple probe design across participants. Intervention involved establishing the missing topographies using a mirror while modeling. Postintervention probes involving learning a new dressing skill from demonstration showed that the establishment of GI across multiple topographies resulted in a decrease in the number of sessions-to-criterion as well as a shift in responding during instruction from emulation to imitation for all four participants. Findings suggest that the establishment of GI across multiple topographies resulted in a new and more efficient way of learning multiple-step chained responses by observing a model in a mirror. Our results provide empirical evidence that GI meets the definition of a behavioral developmental cusp.