In a number of papers1, Castafieda has argued that there is an important distinction between genuine self-belief and belief about something which is in fact oneself. The distinction emerges in our attributions of belief: 'Jake believes that he (himself) is healthy' cannot be analyzed as 'Jake believes that x is healthy', where 'x' is replaced by any name, description or demonstrative pronoun referring to Jake. The reason for the unanalyzability is that genuine self-belief, as opposed to belief about something which happens to be oneself, requires that the believer be able to make first-person reference to himself. A person who did not refer to himself in the first-person way (normally, in English, by use of the pronoun T), could be said to believe of someone, who in fact is himself, that that person is healthy;but such a person could not be said to believe that he (himself) is healthy.
We propose to show how these results of Castafieda's can be incorporated into the formal semantics which Tyler Burge 2 has begun to develop for demonstrative constructions. Since, as we shall argue, attributions of genuine self-belief are best construed as attributions of de re belief, it is useful first to integrate Burge's views on de re belief with his work on demonstrative constructions.
Demonstrative constructions are used to pick out an object extralinguistically without specifying it in the immediate discourse. Formally, this comes to representing the demonstrative construction as an indexed free variable and the entire assertion as an open sentence. Clearly, the satisfaction conditions for closed sentences cannot take into account the peculiar contextual uniqueness of demonstrative constructions. On the other hand, open sentences do account for these contextual characteristics. If someone uses a demonstrative construction to refer to an object, it is as if that person picks