OK, this (Surprising Florida Presidential Election Results) is story about the 2004 Florida results broken down by type of machine in use (Optical scan vs. E-Touch), number of registered Dems and Rep, the expected percentage of votes for each candidate and the actual percentage from the results. It's complicated but really worth looking at. And the results aren't what you might think: The figures show problems with the optical scan machines, not the E-Touch machines. (IMHO, the data in the charts is laid out backwards, making them a little hard to easily read.)

Also look at Brad's Blog here for some additional info. This may just be a strange statistical fluke but it is very consistent as reported in these stories.

I also want to reiterate that I'm NOT suggesting that these anomalies will invalidate the election of Bush (as much as I might wish it would). Investigation into these problems would probably take many months, perhaps years. But I firmly believe problems have to be addressed, thoroughly and openly, to assure the dependabilty and accuracy of the electoral system.