Seriously, how do we know which way undecideds will go?

[BillZBubb]

I've seen the point made here that undecideds generally break against the incumbent on election day. What evidence do we have that this is true?

In the 2000 election, they seemed to break for Gore, who was de facto the incumbent. Same for Bush I in 1988 and Clinton in 1996.

Clearly, undecideds are open to replacing Bush*, but have to be sold that Kerry would be better for them (not necessarily the country). They are also uninformed and apathetic; otherwise, they could not be undecided in this election.

Given that, it appears to me that the undecideds {who even bother to vote) are more likely to stick with "the devil they know" than change to Kerry.

I'd be interested in hearing other theories.