In fact, it is so interesting that Steven Den Beste has broken his vow of silence to post about it. A commenter at Vodkapundit suggests that it simply indicates a non-linear system, but as someone with a little bit of engineering background I would say that non-linearity is cause for further analysis. A good place to start would be to look at a similar graph separately for each of the contributing polls. Much could be determined by whether the anomaly appears consistantly across the spectrum.
Posted by triticale at October 16, 2004 11:49 PM