Research on video modeling has typically utilized either competent peer models or self-models engaging in criterion performances. Although both methods have demonstrated utility in achieving skill acquisition, each has potential disadvantages. The current research utilized a multiple probe design across tasks and replicated across participants in order to demonstrate the efficacy of an instructional video modeling technique to teach functional living skills to three children with autism. Five tasks were selected. Prior to the development of each training video, task analyses were created. Videotapes were developed from the participant's viewing perspective, that is, as the participant would be viewing the task. Instructional video modeling was effective in promoting skill acquisition across all three children and maintained during the postvideo phase and a 1-month follow up.