Time and Newsweek both registered massive bounces for Bush during the Republican National Convention. Rasmussen says he would be showing Bush with a five point bounce (and, therefore, a four point lead) except his Saturday sample was terrible for Bush, giving him a slight 1.2 percent lead in the three day moving average. Now Gallup is showing a two point bounce based on a weekend poll that's moved Bush from one point behind to one point ahead (and now we're in the territory where sampling error matters, so it's not entirely clear than anything changed at all). Obviously, something a bit nutty is going on with polls taken on, say, Friday showing dramatically different results from polls taken over the weekend. Is this "faster public opinion" where people love Bush after seeing his speak and then forget all about it after 36 hours of hurricane coverage? As I recall something similarly screwy happened with Kerry -- he got a big bounce on Friday and then by the following Monday it was gone.
It's hard for me to understand the psychology of folks who would let their votes be swung by a speech -- we've had four years to watch Bush and his performance in office seems like an infinitely better guide to what you should do than is a speech -- so from my point of view there's really no telling.