New Marquette poll: Tommy Thompson 28%, Eric Hovde 20%, Mark Neumann 18%, Jeff Fitzgerald 13%.

With 21% still undecided, the Thompson lead isn't very impressive. And Thompson has gone down since the last poll, last month, when he was at 35%. Hovde too is down. He was at 23% last month. It's Neumann who's up — 8 points from last month's 10%. And Fitzgerald is still at the bottom, but he's more than doubled his support, having been way down at 6% in July.

Although Hovde is down 3 points, the gap between him and Thompson is smaller — from 12 to 8 points.

When undecided voters are asked which candidate they lean towards, the vote becomes 33 percent for Thompson, 24 percent for Hovde, 21 percent for Neumann and 15 percent for Fitzgerald. Seven percent remain undecided. The Republican primary results are based on 519 likely voters (i.e., those who say they are certain they will vote in the August 14 primary).

I see 2 big questions: 1. Is non-Tommy sentiment essentially anti-Tommy? 2. If so, can the anti-Tommy people settle on one of the non-Tommys?

Among the 19 percent of likely primary voters who describe themselves as “very conservative,” there is a close-packed tie for the lead, with Hovde at 24 percent, Neumann at 22 percent, Thompson at 21 percent and Fitzgerald at 15 percent. Among those describing themselves as “conservative,” who make up 52 percent of likely primary voters, Thompson has an advantage at 27 percent to Hovde at 20 percent, Neumann at 19 percent and Fitzgerald at 13 percent. Among the 20 percent of likely voters calling themselves “moderate,” Thompson receives 34 percent to Hovde’s 18 percent, with Neumann at 14 percent and Fitzgerald at 13 percent.

It's hard to figure out how to vote strategically — assuming your goal, as a GOP primary voter, is, above all, for the Senate seat to go to the Republican. But it's an open primary, and Democrats might try to get the weakest candidate in. (But who would that be? Neumann?) Or Democrats might pick Thompson, on the theory that he's the least conservative. The Marquette pollster says that including only Republicans made little difference in the numbers — maybe because Democrats looking at Republicans split between the best loser (Neumann?) or the least-bad winner (Thompson).