This proposal will evaluate a series of peer-mediated interventions (PMIs) for preschool children (3 to 6 years) with ASD and limited or no spoken language, using an innovative Sequential Multiple Assignment Randomized Trial (SMART) design. Available evidence supports the beneficial effects of PMIs for improving social communication in children with ASD. Peer-related social competence is vital to a wide range of child outcomes, such as improved communication and fewer behavioral problems. Unfortunately, approximately 30% of children with ASD remain minimally-verbal in kindergarten, restricting participation in inclusive activities. Recent studies report improved communication after a speech-generating device (SGD) is included in treatment. Effective interventions that can be modified is necessary to ensure optimal communication outcomes when children do not make anticipated progress. A strength of the study is that these interventions can be adopted by community-based, early service providers. All participants will receive an adapted Stay-Play-Talk (SPT) peer-mediated intervention that varies in active ingredients. With SMART designs, it is possible to test and identify alternative combinations of PMI approaches, such as the addition of a SGD. In this study, 132 preschoolers with ASD (and N=264 peers without disabilities) will be initially randomized to SPT and SGD with spoken peer input only (SPT Basic; peers taught to model language) or SPT and SGD with augmented peer input (SPT Plus; peers taught to use verbal language models concurrently with the SGD). Each child's response to treatment after 5 weeks will determine that child's next phase in the SMART design. Children showing a positive response will continue in their originally assigned group; slow responders will be randomly assigned to receive added treatment components to improve communication (either SPT Plus or SPT Advanced). SPT Advanced adds direct instruction strategies (i.e., adult prompts, reinforcers, and teaching trials) to increase child vocalizations in SGD interventions. The use of a SMART design extends our prior work by testing the systematic addition of selected peer-mediated strategies in combination with an SGD that allows for flexible application of interventions based on child response. The investigators have assembled an outstanding team of highly qualified investigators with complementary skills in preschool assessment, language intervention, clinical trials, and statistics.