The manufacturer, Diebold Election Systems, now Premier Election Solutions, disputes the conclusion. Nevertheless, in 2007, the Maryland General Assembly voted to move back to paper ballots, although it will still use e-voting machines in Tuesday's election.

More recently, in a review commissioned by the state of California, researchers at the University of California found that electronic voting machines used in that state had security issues that made them vulnerable to vote tampering. The report prompted California to require that all voting machines also produce a paper trail.

Wallach of Rice University says that touch screens can often be poorly calibrated, causing the on-screen image to be misaligned with the touch-sensitive layer of the screen. Even a properly calibrated machine may not work well for an especially tall or short person because of their angle of view, he adds.

Critics' greatest concern about electronic voting machines, however, is that they might be vulnerable to fraud. "I think it's the complexity and the lack of transparency," says Steven J. Murdoch, a computer security researcher at the University of Cambridge. "It's certainly not apparent to the ordinary voter how it works, or whether it can be tampered with."

Murdoch thinks the move towards electronic voting was driven in part by "modernization for modernization's sake. When I was calling this a bad idea, I was being called a Luddite, but I've spent most of my life working with computers."

However, David Beirne, executive director of the Election Technology Council, which represents voting-machine manufacturers, says that electronic systems are designed to solve real problems. He says that paper ballots are expensive, cumbersome, and often "spoiled" by voters who mark them incorrectly.

"Unfortunately, I think the criticisms have reached such a point that no voting system can satisfy the critics," Beirne says. He also complains that critics often present unlikely scenarios, or ones that could easily be defended against with good management practices. And they don't compare the machines against the vulnerabilities of paper ballots.

Charles Stewart, a professor of political science at MIT who has studied voting technology, agrees that paper ballots are also vulnerable to fraud. "Right now, we know a hundred different ways to corrupt paper systems that any idiot could perform. I don't know of anything that any idiot could perform on voting machines," he says. But Stewart also argues that the industry has been slow to address real security concerns that have been apparent for years.

A number of technological schemes have been suggested for fixing security problems related to electronic voting. The most common is to require that each machine generate a voter-verified paper ballot and to audit a sample of paper ballots after an election. Some states (including California) have moved towards this method.

Another proposal is to use encryption to ensure voters and observers that votes haven't been tampered with. In one such scheme, developed by Wallach and colleagues and called VoteBox, when voters completed a ballot, their identity and a record of their vote would immediately be encrypted and posted online. Each machine would also issue an encryption key to voters so that the record could be decrypted to make sure the vote had been recorded correctly.