People often make ascriptions that they believe to be literally false. A robot, for example, may be treated as if it were a dog, or as if it had certain intentions, emotions, or personality traits. How can one do this while also believing that robots cannot really have such traits? In this paper we explore how Kendall Walton’s theory of make-believe might account for this apparent paradox. We propose several extensions to Walton’s theory, some implications for how we make attributions and use mental models, and an informal account of human-robot interaction from the human’s perspective.