We provide a cognitive analysis of how children represent belief using mental files. We explain why children who pass the false belief test are not aware of the intensionality of belief. Fifty-one 3- to 7-year old children were familiarized with a dual object, e.g., a ball that rattles and is described as a rattle. They observed how a puppet agent witnessed the ball being put into box 1. In the agents absence the ball was taken from box 1, the child was reminded of it being a rattle, and emphasising its being a rattle it was put back into box 1. Then the agent returned, the object was hidden in the experimenters hands and removed from box 1, described as a “rattle,” and transferred to box 2. Children who passed false belief had no problem saying where the puppet would look for the ball. However, in a different condition in which the agent was also shown that the ball was a rattle they erroneously said that the agent would look for the ball in box 1, ignoring the agents knowledge of the identity of rattle and ball. Their problems cease with their mastery of second-order beliefs (she thinks she knows). Problems also vanish when the ball is described not as a rattle but as a thing that rattles. We describe how our theory can account for these data as well as all other relevant data in the literature.