Maybe it's just the heat, but something happening in Iowa. Two Iowa polls of solid methodology -- last week's Post/ABC and today's, released by the University of Iowa, show a virtual dead heat among the Democratic presidential candidates.

Clinton does better when the sample is enlarged to include non just likely caucus voters but Dems statewide, proof for the theory that she has the largest available base. John Edwards maintains a within-the-margin lead among likely caucus goers. His support is dropping.

Why? One possibility is that Iowa Democrats are beginning to see the race as a contest between, increasingly, Obama and Clinton, with Edwards on the verge of losing his status as a frontrunning contender.

The Clinton-Obama exchanges have dominated the last two debates and certainly have accounted for most of the press coverage. Obama might be helped by his steady television ad schedule in Iowa, too. Clinton, who has seen her lead in most national polls rise a bit, isn't on the air at all, so her campaign has every reason to treat these latest two polls as a sign of strength.

There's more good news for Bill Richardson: he continues his slow climb.

Obama seems to have Mitt Romney's problem. To wit: his standing in national polling is going southward and his standing in key state polls is trending in the opposite direction. The audiences are different; Democrats nationwide, who rely mostly on national press coverage for their information these days, could be forgiven if they thought Obama had a rough couple of weeks. The press thought his foreign policy answers were green, and his own ties to lobbyists have come under heavy scrutiny.

Democrats in Iowa, who (in theory) are paying somewhat closer attention and who've gotten to see the candidate during this period, might look past the press coverage and find they like the contrasts Obama is drawing.

But his rise is not coming at the expense of Clinton's fall.

** The only dynamic worth paying attention to in national polling right now is trends; all other things being equal, a trend has some significance; the data points themselves have virtually none at all.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.