For children, high quality friendships are associated with adaptive social, emotional and academic functioning. There is also evidence that children experience real and imaginary friendships in similar ways, and that imagined relationships could have an impact on development. However, less is known about the relationships made possible by virtual agents in digital media. This dissertation research was designed to provide preliminary data about children’s concepts of virtual agents, and the social opportunities they attribute to such entities. In Studies 1 and 2 (combined N = 48), preschool aged children differentiated the social affordances of a stuffed dog and a virtual dog. Participants played a game in which they guessed whether a child in a video was referring to a stuffed dog or a virtual dog in a series of statements. Items designed to assess high quality friendships, such as comfort, protection and love, were attributed more to the stuffed dog than the virtual dog. Studies 3 and 4 examined adult and child concepts of a virtual child, and how concepts of this entity might differ from a real child, a child on a video chat program (e.g., Skype™) and an inanimate doll. Adults and children attributed a range of properties to each child agent, including biological, psychological and social properties, as well as opportunities for relationships. In Study 3 (N = 144), adults did not differentiate between the virtual child and the doll on the social property; however, they favored the doll on opportunities for unilateral relationships. In Study 4 (N = 30), five to eight-year-old children indicated an overall preference for the doll on the social property, as well as on opportunities for reciprocal relationships. Children also favored the doll on opportunities for love, companionship, and intimate disclosure. Altogether, these findings suggest that virtual agents afford more limited social opportunities than inanimate artifacts, and they are less likely to be loved by children and adults alike. These results raise important questions about the design goals for virtual agents, and the functions they are intended to serve in our everyday lives. This dissertation includes both previously published and co-authored material.