A total of 104 children aged between 41 and 47 months were selected to study the relationship between language and false belief understanding. Participants were assigned to four different training conditions: discourse, labelling, control (all with deceptive objects), and sentential complements (involving non-deceptive objects). Post-test results showed an improvement in children"s false belief understanding in the discourse and the labelling conditions, but not in the sentential complements with non-deceptive objects or the control group. Furthermore, the most remarkable improvement in false belief understanding occurred in the labelling group. These results suggest that some types of linguistic experience promote the development of false belief understanding, provided that differing perspectives are confronted.