In Study 1, we presented children with a familiar object (e.g., a screwdriver) that was given a novel label (‘‘sprock’’) and a novel function (pushing a ball out of a tube) in either a pretend or real context. Children were then (a) shown a set of objects and asked to identify which object in the set might be a real sprock and (b) asked to demonstrate what sprocks do. The objects varied in terms of how similar they were to the screwdriver in appearance and function. If children do not transfer any information from pretense to reality, children in the pretend condition would choose at chance when asked to select a real sprock and would not demonstrate a function learned in pretense once the pretend episode was over. In contrast, if children do recall a novel function they were taught during pretend play, this would suggest that they can transfer function information from pretend to reality. In the study, 5-year-olds were tested because they are in the ‘‘high season’’ of pretend play (Singer & Singer, 1990) and thus, might be most likely to learn in pretend contexts.