I hate the "Is Bush likely to win" question.
It is misleading. Zogby asks if Bush is very likely, somewhat likely, or not likely to win. Then, he combines the somewhat and very likely #'s to get that 60%. I think that's a crock.
If the #'s broke down this way:
Bush will win?
Very likely: 20
Somewhat likely: 40
Not likely: 37
You could just as easily say that "77% of those polled had little confidence in Bush's reelection chances, giving him a "not likely" or "somewhat likely" chance.
Also, since "somewhat likely" is the middle option of 3, it should be seen as neutral and realistic, not pessimistic.