The year was 1984, and the state was Iowa. A white man who had just voted walked out of his precinct caucus and saw the Rev. Jesse Jackson standing outside.
"I did all I could," the man told Jackson ruefully, "but I just couldn't bring myself to pull the lever and vote for you."
L. Douglas Wilder laughs as he relates the story Jackson once told him, the sting eased by time and Wilder's vantage point as the nation's first elected black governor.
Now it's a quarter of a century later, and the man everyone's talking about is Barack Obama, the Illinois senator holding a slim lead in many polls. But can the polls be trusted? A central question about race and politics hasn't changed since 1984: Do white people lie — to pollsters or even to themselves — about their willingness to vote for black candidates?
In the not-so-distant past, the consensus was a clear yes. Today, however, there is widespread disagreement about whether Obama is subject to the predicament known as the Wilder or Bradley Effect — whether in the privacy of the voting booth, white people will actually pull the lever for the first black man to come within shouting distance of the presidency.
Given that surveys can have trouble uncovering the truth about many things besides race, plus the massive technological, demographic and cultural changes in play, this question is contributing to an almost unprecedented air of uncertainty surrounding this year's polls.
In 1989, Wilder polled as many as 15 points ahead in the days before the election for Virginia governor, but squeaked into office by a minuscule 6,700 votes. David Dinkins had a similar experience that year, when he became New York City's first black mayor. And the phenomenon was first noted in 1982, when Tom Bradley endured a stunning defeat in the California governor's race after exit polls indicated he was the winner.
The reason for these disparities? A significant amount of white people did not admit that race played a role in their voting decision, pollsters and academics say. Another factor: When the person asking the questions was black, respondents were more likely to say they favored the black candidate.
In the recent Democratic primary, exit polls in 28 states overstated Obama's actual share of the final vote.
Andrew Kohut, president of the Pew Research Center, doesn't think people are lying to pollsters today about their support for Obama, "because I don't think there's a lot of stigma in saying you're voting for John McCain." Kohut said it's not like polls are asking, "Do you want to vote for the white guy or the black guy?"
But he did see potential for error based on the people who decline to participate in polls, whom he describes as largely lower-income whites more likely than the population at large to have racially intolerant views.
"The real frailty of our polls is that we get very high refusal rates, and we survive because the people who we interview are like the people who we don't interview on most things," Kohut said. "(Racism) is not one of them."
So are current polls accurate? "I don't know," Kohut said, "and to be honest with you, this is something every pollster I know is concerned about."
Wilder, now the mayor of Richmond, Va., said his internal polls during the governor's race showed it to be much closer than most people thought. "It was clear that people were having the first opportunity to vote for an African-American, and there was uncertainty," he said. "You know, 'Is he going to be fair, is he just going to look out for his own people. And who are his own people?' I think we've come a great distance from that. I've seen the progress."
So is Wilder ready to bury the Wilder Effect?
"No, I won't say that," he said with a laugh. "I won't go that far."
Daniel J. Hopkins will. The Harvard University postdoctoral fellow examined data from 133 gubernatorial and Senate elections from 1986 to 2006 and concluded that the effect vanished in the early 1990s as racially divisive issues such as crime and welfare reform receded from the national stage.
Hopkins said that race could play a larger role if it is injected into the campaign — as it often is in the waning days of close contests involving black candidates.
Days before the 2006 Senate election in Tennessee, with polls showing the race almost deadlocked, Republicans released an ad featuring a ditzy blond actress saying she met Harold Ford Jr. at the Playboy Club and asking the black Democrat to "call me." Ford lost.
In 1990's tight North Carolina Senate contest, Republican Jesse Helms was running about even with Democrat Harvey Gantt when he released an ad showing white hands crumpling a job rejection letter as a narrator mentioned racial quotas. Helms won.
Blacks, too, have sought to use race to their political advantage: In a congressional primary this month in Memphis, a black challenger tried to link the incumbent, Steve Cohen, to the Ku Klux Klan. Cohen won easily.
While Obama may face some of these historical hurdles, there are other, unprecedented factors at work: a presidential instead of statewide election, a spike in black voters and the increase in young voters who are more racially tolerant, watch more YouTube than television and eschew the land telephone lines used by most polls.
The racial pendulum may even have swung back the other way, said Anthony G. Greenwald, a psychology professor at the University of Washington, citing a "reverse Bradley Effect" during the Democratic primary: In states with larger black populations, such as Virginia, South Carolina, Mississippi and Georgia, Obama got more votes than polls predicted.
Like Kohut, Greenwald doesn't think people are deliberately lying in polls. But he does see potential for polling errors due to undecided white voters overstating their support for Obama or choosing McCain at the last minute, and the influence of "racial attitudes and stereotypes that people in many cases are not aware they have."
Many pollsters are trying to adjust their methods to account for these unprecedented variables. It's not easy, however, to solve these new problems in the heat of a tight presidential race.
"I don't think anyone is correct or incorrect, including me," Greenwald said of the current poll numbers. "To get to the heart of that, you'd have to do the kinds of research that haven't been done."
The Obama campaign declined to comment on how it conducts its polling. The McCain campaign did not respond to requests for comment.
Matthew Dowd, an ABC News commentator and former chief strategist for President Bush's 2004 campaign, expects the Wilder Effect to be a "small factor" in November. "I wouldn't want to be Barack Obama and up two points going into Election Day," he said.
"My guess is that (the Obama campaign) understands that and they know it's not enough to be ahead," Dowd said. "They have to be ahead by a lot."