Video-modeling (VM) is a widely used instructional technique that has been applied to the education of children with developmental disabilities. One form of VM that lacks in-depth analysis is point-of-view video modeling (POVM). The current study investigated the use of POVM to teach three children diagnosed with autism to initiate and maintain a conversation with a conversant. Using a multiple baseline across scripts design, the participants were taught to engage in both eye contact and vocal behavior without the presentation of a vocal discriminative stimulus from the conversant. The treatment package included both the presentation of the target video and reinforcement of target behavior. Although this combination proved successful for increasing the social behavior of two participants, prompts were necessary to achieve acquisition for a third. These data suggest that while POVM may be a successful technique for teaching some social skills, limitations exist that should be further investigated. (Contains 2 tables and 7 figures.)