Perry on Frege on Demonstratives

John Perry's argument about the irrelevancy of belief to the analysis of demonstratives (from "Frege on Demonstratives"):

1. Suppose that if a speaker S uses a demonstrative (e.g. today, here, this, etc.), then S has to have a descriptive belief that uniquely identifies the referent of the demonstrative.
   
2. Consider the following example: On 9/12/01 at 12:30 a.m., S looks at his wife and says, "Today the whole world changed". 
3. By (1), S must have a descriptive belief that uniquely identifies 9/12/01.
4. But S's beliefs in the vicinity identify only 9/11/01. 
5. So (1) is false. 

This was one of Perry's many arguments that, taken together, seem to show that Frege's philosophy of language can't handle demonstrative expressions.