Background The aim of this study was to assess the effectiveness of two communication interventions for preschool-aged children with autism. Method Six typically developing peers were taught to implement peer-mediated naturalistic teaching, with and without a speech generating device (SGD), during play sessions with 3 classmates with autism in three preschools. Generalisation probes were conducted during mealtimes at the preschools. A multiple baseline design was used to assess the outcomes of the two intervention conditions. Results All 3 children with autism increased their communicative behaviours immediately following the introduction of the two interventions, and generalised these increases to mealtime interactions with their peers. However, only 1 child maintained these increases in communication. Conclusion These results provide preliminary evidence for the effectiveness of combining peer-mediated naturalistic teaching with the use of SGDs for preschool-aged children with autism. Suggestions for improving the maintenance of intervention effects are provided.