In November's midterm elections, Buchanan edged out Democratic challenger Christine Jennings in the race for Florida's Congressional District 13 by a slim margin of 369 votes, according to the election results certified by the state. However, Jennings and some voters in Sarasota County, the most populous county in the congressional district, have filed lawsuits contesting the results.

The problem? A statistically improbable number of people in the pro-Jennings county--about 18,000 or 13 percent of all voters--failed to register a choice in the race. An undervote--as such failures are dubbed--seldom occurs in those numbers. While many people decide not to vote for any candidate in a race, high rates of undervoting usually happen only in the less publicized races. Voters fail to vote in major races generally far less than 5 percent of the time, according to voting experts.

Moreover, voters that did not use the county's election machines had a typical chance of not voting in the race, failing to make a selection only 2.5 percent of the time. On the e-voting machines used by the county, however, the story was different: Nearly 15 percent of the people who used the iVotronic systems manufactured by Election Systems & Software failed to vote in the Representative race.

"Both sides agree that thousands of votes were not counted and that those lost votes changed the outcome of this race," Kendall Coffey, attorney for the Jennings campaign, said in a statement. “We need to find out exactly what suppressed those votes, and we feel that the rights of Florida’s voters to have their votes count – and be counted accurately – is paramount in this case."

It's a result that even the winner could not dismiss. In a statement issued last week, Buchanan asked Jennings to acknowledge that he won the election but noted that the evidence did point to problems with the race.

"As with the 2000 presidential outcome, it is likely to go down as another botched election due to a combination of flawed ballot design, voter error and voter protest," Buchanan stated. "With no paper trail, it is impossible to discern the precise reason for the 18,000 Sarasota non-votes in this race. It is an unknown that is impossible to discover."

E-voting machines have been the target of a great deal of criticism, in part because the technology has been adopted quickly in an attempt to solve problems seen in the 2000 election with punchcard voting machines, but also because electronic voting systems are generally based on a personal computer architecture and use software that has not adequately been tested for security vulnerabilities. While many detractors of the technology focus on the danger of fraud posed by the machines, actual election day complaints regarding the systems have to date amounted to mechanical defects, software failures, or administration problems.

While such problems create distrust among voters, it's the more systematic issues of secretive certification, poor security checking and the lack of a software-independent method of auditing that concerns experts. The New York Times reported on Thursday that the federal government had blocked Ciber, a major e-voting system certification lab, from certifying any more systems as of last summer. Security researchers have repeated pointed out flaws in the system software of e-voting machines, especially those manufactured by Diebold Election Systems.

Last month, the National Institute of Standards and Technology, government agency that recommends technological standards, told election administrators that states should not rely on electronic voting machines that have no way to independently verify the tally. The report moved away from discussing specific requirements, such as paper audit trails, and instead couches security concerns in terms of the dependence of vote verification on a system's software.

"I thought they hit the nail right on the head," Avi Rubin, computer science professor at Johns Hopkins University and the director of A Center for Correct, Usable, Reliable, Auditable, and Transparent Elections (ACCURATE), said in an interview at the time. "I think software independence is the right metric to use when discussing a voting system."

In the Sarasota election, however, the machines are not likely the primary reason for the massive undervote, experts said.

Political science and voting-system experts believe they know exactly what happened.

A group of seven e-voting experts hired by the state of Florida to review the case and check iVotronic's software will likely conclude that a poorly designed ballot and a lack of a meaningful undervote warning on the Election Systems & Software machine resulted in many voters missing the Congressional race, a source familiar with the group's deliberations told SecurityFocus.

The theory is backed by a 60-page statistical analysis penned by four political-science researchers that concluded that, if the county's voters had used the election equipment and ballot design that were used by other counties, Jennings would have received at least another 3,000 votes and won the election.

"I think that one thing this election will do is show that we need to study to how best to represent a ballot," said Michael Herron, professor of government at Dartmouth College and one of the authors of the analysis.

In the paper (PDF), Herron and his colleagues used statistical analysis to show that the undervote only happened in three counties: for the U.S. House of Representatives' race in Sarasota and for the Attorney General's race in Charlotte and Lee counties. In all three cases, the race with the large number of undervotes was paired with another election contest on the same page.

Putting a single race on each page could solve the undervote problem, but having a more obvious warning to the voter that they missed a race could be a better solution. ES&S's iVotronic does not have an immediate undervote warning but does show that the voter did not make a selection in a review screen at the end of the process.

"Every iVotronic has a review screen giving voters a chance to check their choices and see if they missed a race," said Ken Fields, a spokesperson for the voting system maker.

The group of seven voting experts hired by the state of Florida, and led by Alec Yasinsac, an associate professor of computer science at Florida State University, will likely conclude that the ballot review at the end of the voting process is not enough to prevent undervotes, according to a source familiar with the deliberations of the group. So far, no major flaws in ES&S's software have been found that could have affected the election, said the source, who asked not to be named.

The fight may have ended for all practical purposes, but the Jennings campaign has not given up. Jennings legal team and another group of voters appealed a circuit court decision last week, after ES&S and Florida officials successfully prevented the challengers from winning an order to open up the voting machine maker's source code.

Finding that the election problems were primarily caused by ballot design will likely mean that the result will stand, said Dan Tokaji, a professor of law at the University of Ohio's Moritz School of Law. Since there is no way to determine how the people who missed voting would have voted, the only way to change the election would be to have a revote--an unlikely outcome, he said.

"If it was a software problem, then I think the argument for a revote is much stronger, but if it because (voters) missed the race, then I think that a revote is much more unlikely," Tokaji said. "There are problems and glitches in every election, so we have to ask ourselves the question that in what circumstances are the problems so serious that we can't stand them and need to have a new election."

Looking beyond the election, it's not clear what the best solution would be to the undervote problem.

A dialog box box that pops up to warn a person when they have not voted in a race could become a distraction, Tokaji said. "When we get to the point of having dialog boxes to warn of undervotes, I worry that we are going to slow down the voting process," he said.

Forcing voters to vote in every race, but giving them another choice--none of the above--could also solve the problem, but could also take more time. The strategy also raises the fear among politicians that a majority of voters could choose "none of the above," essentially casting a vote of no confidence in the candidates.

However, it's a choice that Florida's current winner has now also proposed.

"State legislators should put an end to the mystery of 'undervotes' by requiring a 'none of the above' box be included for all races," Rep. Buchanan said in last week's statement.

"That would render a ballot invalid if a choice weren't made in all races, and poll workers would spot the omission and alert voters immediately so they could correct it," he said. "There would be no more frustrated voters who fear their votes weren't counted."