I've been trying to imagine what would impel a judge to order the state to put 12 dead people back on the voting rolls after they'd been removed, yet that's what U.S. District Judge John Kane did last week in an attempt to show Secretary of State Mike Coffman who's boss.

Now, maybe you thought the integrity of an election depends in part on getting dead people off the rolls. If so, that just shows how out of step you are with modern election management.

The judge also ordered 66 duplicate registrations back on the rolls, as well as one felon, one noncitizen, four voters who'd asked for their registration to be purged and 62 who'd moved out of a county or state, according to Coffman. Fortunately, Coffman had to nullify only two days' work by election officials - the period after Kane approved a truce between Coffman and activist groups who dog his every step and who would have little trouble finding a reason to keep a Great Dane on the voter rolls.

If Kane's ban on cleaning those rolls had been in place all summer and fall, on the other hand, there'd be thousands of dead and duplicate voters registered, not to mention thousands more who'd moved. Talk about an invitation to fraud.

Make no mistake: Most of these registration problems occur naturally or by accident, not as a result of would-be scam artists. Coffman suggests, for example, that many duplicates involve people who aren't sure they are registered because they don't always vote and so sign up again when approached by a volunteer with a clipboard.

Still, if thousands of mail ballots had been sent to people whose records should have been removed, at least some of those ballots would have been filled out and returned under false pretenses - and then counted.

Would the cheaters have been caught? In some cases, absolutely. In others, probably not.

And let's not even speculate about the possible effect on razor-tight races.

The puzzle of polling

A week ago, the Rocky Mountain News/CBS4 poll showed Amendment 58 - that's the one that would hike severance taxes - with the support of 49 percent of voters. Five days later, a poll for The Denver Post showed the same amendment with only 36 percent support.

The two polls - both by reputable firms - came to jarringly different conclusions on several other ballot measures, too, such as Amendments 46 and 59.

Did voter attitudes change so much in a few days? Not likely. But public opinion is notoriously difficult to measure, even on something as simple as an up or down vote.

Still, election polls have a pretty good track record. It's the nonelection surveys we ought to worry most about, as my daughter recently reminded me. She'd just read Herbert Asher's Polling and the Public for a college course, and it left her understandably skeptical of pollsters' attempts to measure opinion on complex matters.

It's bad enough that the content of some polls is stacked by sponsoring groups to nudge them toward a predetermined result. Far more disturbing is the extent to which seemingly innocuous word choice by conscientious pollsters can result in huge swings in opinion - the use of "aid to the needy" in a question, for example, instead of "welfare."

In one revealing example recounted by Asher, nearly 59 percent of Americans surveyed said Bill Clinton should resign if impeached rather than "fight the charges in the Senate" but only 43 percent said he should resign when the alternative was "remain in office and face trial in the Senate." Who could have guessed?

Come next year, pollsters are going to tell us how Americans supposedly feel about the president's plans on everything from health care and energy to taxes - and lots of other issues that most people hardly follow at all. Yet if pollsters can't agree even on the level of support for a publicity saturated amendment such as 58, why should we believe them when they tell us, say, what percentage of Americans favors a system allowing utilities to buy and sell carbon permits?

Vincent Carroll is editor of the editorial pages. Reach him at carrollv@RockyMountainNews.com.