The current study used a training methodology to determine whether different kinds of linguistic interaction play a causal role in children's development of false belief understanding. After 3 training sessions, 3-year-old children improved their false belief understanding both in a training condition involving perspective-shifting discourse about deceptive objects (without mental state terms) and in a condition in which sentential complement syntax was used (without deceptive objects). Children did not improve in a condition in which they were exposed to deceptive objects without accompanying language. Children showed most improvement in a condition using both perspective-shifting discourse and sentential complement syntax, suggesting that each of these types of linguistic experience plays an independent role in the ontogeny of false belief understanding.