I tested the effects of teaching imitation using a mirror on the emergence of generalized imitation (GI) of physical movements by children with autism. I first tested if 128 adults, from 19 to 56 years old, emitted mirrored or non-mirrored responses in GI, because this had been a point of theoretical debate and little data were available. I found that typical adults emitted both mirrored and non-mirrored responses during the GI probe. Accordingly, topographically correct non-mirrored or mirrored responses were considered correct for Experiment Two. In the second experiment, using a combined experimental-control group design with a "nested" non-concurrent multiple probe design across participants, I compared a mirror-trained group (3 children, ages 3 to 4 years olds) and a non-mirror (face-to-face) trained group (3 children, ages 3 to 4 years olds), controlling for the amount of instruction and maturation. The dependent variable was the numbers of correct untaught imitative responses during face-to-face pre and post intervention probe sessions, and the independent variable was the mirror training.

The results show that the mirror instruction facilitated in the emergence of GI in all 3 children with autism in the mirror-trained group: their controls that received the same amount of instruction face-to -face did not. I concluded that mirror training resulted in reinforcement of the duplication production and non-mirror training reinforced individual movements.