But the detail is far more interesting, and prognosticators should read the caveats at the end of the poll carefully. Merv Field is known for his honesty in describing the shortcomings of polling, and I have never seen Merv and his colleagues as cautious in a description as they are in this one. Also note how small the sample is, how small the subgroups are and that most of the telephoning for the survey was done before the Democratic debate last Thursday night. I’d pay attention to the overall trends expressed / young vs old, Latino vs. African American preferences, gender gap etc. among Democrats. I’d be much more cautious, as indeed the poll itself seems to be in the fine print, about the head to head between Clinton and Obama and the assumptions about who is a likely voter. With so many independents in the mix, there is a good chance the assumptions from past years don’t work.

The other really interesting element is the huge number of undecideds – 18 percent – this late in the game.