A bottle of fruit juice is labeled to inform buyers that it is free of GMOs, (or Genetically Modified Organisms), is displayed in Boulder, Colo.. (AP/Brennan Linsley)

Yesterday, the Oregonian released a new poll it had commissioned showing that Measure 92, a ballot initiative to require the labeling of foods containing genetically modified ingredients in the state, was trailing 48 to 42 percent among likely voters. Yet back in July, another poll conducted for Earthfix of Oregon, an environmentally focused branch of Oregon Public Broadcasting, put support for the initiative at a very impressive 77 percent. That's a pretty steep dropoff.

We've heard this story before. Last year, Washington state rejected GMO labeling at the ballot, with just under 49 percent of citizens voting in favor. Before that, in 2012, a labeling initiative failed in California, 48.6 percent to 51.4 percent.  This is despite national polling suggesting that the labeling of genetically modified foods is stunningly popular: A 2013 New York Times national survey, for instance, found that a staggering 93 percent of American support it.

So why does GMO labeling keep losing in actual ballot initiatives -- even in pretty lefty places like California, Washington, and now possibly Oregon?

Those in favor of GMO labeling -- who argue in favor of consumer choice, and also intimate that GMO foods, modified through modern biotechnological techniques, could potentially be dangerous --  will surely say it's because large volumes of food industry dollars are being dumped in these races, to kill the initiatives with huge ad buys. And indeed, according to The Oregonian, the fight over Measure 92 is "the costliest ballot measure fight in Oregon history," with those spending heavily to oppose the initiative including Monsanto and Dupont Pioneer.

But there's another possible explanation -- namely, that when people tell pollsters they favor GMO labeling, they don't really know what they're saying. Because overall public knowledge about GMOs is very low, many GMO polls give "a measure of what people will say they want to label when they have no idea what that means," explains Yale public opinion researcher Dan Kahan.

"I am sure you can replicate that with other samples of people who have no idea what you’re talking about," Kahan adds.

The data supporting this interpretation -- that Americans don't actually know a lot about genetically modified foods, and so polls suggesting they support their labeling should be taken with a major grain of salt -- are fairly compelling. One 2013 survey conducted by researchers at Rutgers University found that 54 percent of Americans say they know “very little or nothing at all” about genetically modified foods, and 25 percent have never even heard of them. Only 26 percent of Americans, meanwhile, were actually aware that GMO labeling is not currently required.

"It’s really clear that people don’t know very much about the subject," says Rutgers' William Hallman, lead researcher on the poll. "And when people don’t know much abut a subject, how you ask them a question about it largely determines the answer you get back."

Indeed, Hallman's survey also found that when you ask people in the abstract, “What information would you like to see on food labels that is not already there?”, most say they don’t want any more information on the label -- and only 7 percent voluntarily come up with GMOs as an answer. So while over 90 percent of Americans may say GMO labeling is a good thing when you actually ask them directly about it, the vast majority of people are not going around thinking that idea independently of being prompted.

So then what happens when on GMO ballot initiative is actually  up for a decision in a given state? First of all, explains John Gastil, a professor at Penn State who studies ballot initiatives, these initiatives generally do worse than initial polls suggest they'll do. "The reason is that fortunately, we have an instinct which tells us, if we don’t understand something, perhaps we should vote against it," says Gastil. "That’s a general phenomenon that many voters use as a heuristic."

And what happens when voters actually get to know the GMO labeling issue, inside out? We actually have intriguing evidence on that.

Oregon has actually created a process called a Citizens' Initiative Review, in which a random sample of 20 citizens hear from both sides of an initiative (and outside experts), and then come up with a report laying out the pro and con case that is then included in the state's voter guide. Such a review was conducted for Measure 92. Pro arguments included "more control and transparency over our food purchasing decisions" and that the initiative "could benefit Oregon family farmers that grow traditional crops by increasing public demand for crops that are not genetically engineered." Con arguments, meanwhile, included the assertion that “Existing food labels already give consumers a more reliable way to choose foods without GE ingredients if that is what they prefer, including 'organic' and 'non-GMO' labels. Measure 92 conflicts with these national labeling standards." In this case, the panel’s ultimate assessment of  Measure 92 split very evenly, with 9 panelists in favor of it and 11 against.

"Even after several days of study, you had voters kind of torn in these matters," observes Gastil. Which is still more evidence that general polls don't reflect how people really think about the issue of genetically modified food labeling -- when they actually think about it.