Two quick notes about the polls:
1) Most analysts, when they describe how the polls might be underestimating McCain's strength, point to the possible "Bradley Effect", which states that certain people who tell pollsters that they are going to vote for Obama, in fact are not, because of racial bias.
It's possible that there will be some Bradley Effect, especially in the kinds of places that Murtha denigrated last week.
However, there may be another effect which is also underestimating McCain's strength. In places with an educated, politically correct electorate, people might have a lightly considered, visceral inclination to back Obama (and thus tell pollsters that they intend to do so), because it seems like the right thing to do. The New York Times sure seems to like him ... I'm going to support him, too! Supporting Obama is the politically correct thing to do. I am quite sure that there are a number of dinner parties on the Upper Westside that people would not get invited to if it were generally known that they were McCain supporters. Again, these are not heavily researched, deeply held convictions, but rather lightly considered, visceral feelings.
But is it not possible that the doubts that McCain has tried to plant in people's minds might actually be working, and some of these people might end up pulling the lever for McCain because, in the end, when they actually get to the voting booth, they view him as the safer candidate? In fact, my gut tells me that this effect could be as large (on national vote totals) as the Bradley Effect, although it will have less of an effect on the election itself, since it would tend to be felt in the blue states rather than red ones and therefore may not effect electoral college math. If big enough though, it could make a difference in certain towns like Philadelphia, Minneapolis, Charlotte, etc, which could be important at the margins.
2) People are simply not paying enough attention to how charged up the Republican base is because of Sarah Palin.
George Bush beat John Kerry, in large part because the conservative base showed up in numbers heretofore not seen in a national election.
This could happen again. Ironically, many of the polls which show large national advantages for Obama do so because they calculate huge voter turnouts from younger, black and hispanic voters.
Wouldn't it be interesting if that effect was cancelled out by the huge, Sarah Palin influenced turnout of the Republican base?