Having established that (a) children must distinguish between pretend and real, (b) there is not complete permeability across these contexts, and yet (c) real information must wend its way from the real world into the pretend world, we ask whether there is selective transfer in the opposite direction such that at least some information crosses from the pretend realm into the real realm? Many studies have investigated whether children will learn novel information from fictional stories, but rela-tively few have investigated the analogous question in pretend play. Two recent studies might support the idea that some information can cross from pretend worlds to real ones, enabling learning from pretense . In both studies, preschoolers were shown a puppet introduced as a ‘‘nerp’’ and then told about the nerp’s preferences and fears. For example, the nerp pretended to eat and enjoy a cherry (represented by a red bead) but pretended to dislike a carrot (an orange bead). Then (to demarcate the pretend and real situations) the experimenter put the puppet away and brought out a book with a photograph of a loris (an animal most children have not seen or heard of). Children were told that the loris was a nerp. For the test, children were asked four questions about what the nerp did and did not like; in some studies the questions were forced-choice, pairing objects seen previously with new objects, and in others they were open-ended. Children performed quite well on the forced-choice questions (e.g., ‘‘Do nerps not like to eat carrots or corn?’’), but across several studies performance on open-ended questions (e.g., ‘‘Can you tell me what nerps do not like to eat?’’) was approximately 50%. Because responses to the forced-choice questions might be due to recognizing what had previously been associated with nerps, the open-ended results suggest that learning from pretense, although possible, may be difficult for young children. In addition, in these studies there was no comparison case of extending from real to real; thus, we do not know how learning in pretense contexts compares with learning similar information in real contexts.