Many children with autism (CWA) demonstrate significant deficits in symbolic play. Previous research has demonstrated that in-vivo modeling and most-to-least prompting have been effective training tools for teaching a variety of play skills to CWA. The purpose of the current study was to (1) examine in-vivo modeling and most-to-least prompting procedures for teaching CWA sequences of play using dress up materials and toy play objects, and (2) to determine whether generalized symbolic play emerged across untrained combinations of dress up materials and toy play objects. During training, teachers modeled a play sequence, then used a total task task analysis and prompting to teach the CWA the play sequence. Once mastery levels were observed in training, teachers replaced trained objects with objects from another play sequence to test for symbolic play. Interobserver agreement was calculated for 50% of sessions ranging from 66.66% to 100% and averaging 93% agreement. Results suggest in-vivo modeling and most-to-least prompting were effective in teaching symbolic play to CWA.