It's a very long way to November. Head-to-head polls are pretty meaningless at this point.
The polls at this point are pretty much useless in determining the outcome in November. If they make you feel better, or justified, or whatever, go ahead and follow them, but just don't make any decisions based on them.
In truth, May horse-race polls have the predictive powers of a 7-year-old dressed up as a swami and using an upside down goldfish bowl to peer into the future. In the last five presidential elections, the Gallup Poll conducted right before Memorial Day got the eventual winner of the national popular vote wrong. From Michael Dukakis' 13-percentage-point lead in 1988 to John Kerry's 4-point edge in 2004, the Gallup Poll (like other national surveys) pointed in precisely the wrong direction. The only unequivocal success story was 1996, when Gallup (along with virtually every sentient American) figured out that the then-youthful Bill Clinton would handily defeat Bob Dole, the oldest first-time presidential nominee in history (McCain included).
What passes as political news in May often looks like trivia in hindsight. Four years ago this week, the political news provided few clues that the 2004 election would pivot around Republican turnout in rural Ohio. Kerry was still bandaging the last lingering wounds from the Democratic primary, effusively praising Dick Gephardt (and implicitly boosting him as a vice-presidential favorite) before the Teamsters in Las Vegas and then appearing with Howard Dean at an antiwar rally in Portland, Ore., designed to counter Ralph Nader's return-bid spoiler presidential campaign. For the record, Gephardt was not Kerry's running mate (though Rupert Murdoch's New York Post ballyhooed his selection in an epic wrong-call front-page headline) and Nader, who did not appear on the ballot in Oregon, won only 464,000 votes nationally.
Salon Article by Walter Shapiro