For weeks for now — and especially after the gun background check bill failed in the Senate last week — gun control activists have been repeatedly invoking the figure of 90 percent, claiming that was the number of people that supported the legislation.

Typical was Washington Post writer Scott Clement who wrote a post headlined: “90 percent of Americans want expanded background checks on guns. Why isn’t this a political slam dunk?”

President Obama was one of the most aggressive users of this factiod. For example, last month he said:

And if you ask most Americans outside of Washington — including many gun owners — some of these ideas, they don’t consider them controversial.  Right now, 90 percent of Americans — 90 percent — support background checks that will keep criminals and people who have been found to be a danger to themselves or others from buying a gun.  More than 80 percent of Republicans agree.  More than 80 percent of gun owners agree.  Think about that.  How often do 90 percent of Americans agree on anything?  (Laughter.)  It never happens.

Here’s a possible solution to this confounding mystery: Maybe the factoid isn’t true. USA Today reports that a new poll it commissioned on gun control found something else entirely:

Americans are more narrowly divided on the issue than in recent months, and backing for a bill has slipped below 50%, the poll finds. By 49%-45%, those surveyed favor Congress passing a new gun-control law. In an NBC/Wall Street Journal poll in early April, 55% had backed a stricter gun law, which was down from 61% in February.

The survey of 1,002 adults was taken Thursday through Sunday by Princeton Survey Research. The margin of error is +/- 4 percentage points.

This is not to say that the earlier polls showing the 90 percent figure were lies — or even wrong at the time they were taken, strictly speaking. Several polls in March showed this,  as did some polls taken earlier this month.

What it shows is that the public opinion on this issue is not fixed. My guess is that when the March polls were conducted — during the height of the debate — many people normally inclined to oppose such measures were cowed by the bloodshed in Connecticut and too embarrassed to say they didn’t support new laws, even to a pollster. That probably faded as became the apparent the bill wasn’t a slam-dunk.

In addition, many of those who did genuinely support such a measure probably didn’t support it very deeply. That is, after Newtown they would be ok with some new regulations, but they weren’t all that personally invested in it.

My father is one of those. A Republican and owner of several guns, he would be fine with expanded background checks but it is not an issue that moves him to write his congressman, march in a protest or anything like that.  By the time the next election rolls around, he’s not likely to take a candidate’s stance on it into consideration one way or another. (Taxes, on the other hand…) And yet if a pollster had called my father he would have been part of the supposed 90 percent.

The debate on gun background checks is a good reminder that even honest polls can mislead. They can tell you how many people support a given issue at that moment in time, but they don’t always tell you how passionately they support that issue — or how likely they are to change their minds.