Voters generally prefer electronic voting machines to paper-based alternatives, but some e-voting machines have error rates of 3 percent or more, according to a study released Friday.

Voters generally were most comfortable with some models of touch-screen e-voting machines, often called direct record electronic (DRE) machines, when tested against paper ballots and e-voting machines using buttons and dials, said the study, published by the Brookings Institute, a centrist think tank.

In five DRE systems researchers tested, the error rate of the worst-performing machines was 3 percent in a simple task such as voting for president, researchers said. In more complex races, the error rate, the rate at which voters voted for the wrong candidate, was higher. Researchers urged voting machine manufacturers and elections officials to focus more on ballot design, saying badly designed ballots caused many of the problems.

"You might think, 'Hey, a 3 percent error rate, that's pretty good,'" said Paul Herrnson, a political science professor at the University of Maryland and lead author of the study. "But ... 3 percent is not good enough in an election, because it can change the outcome. This shows us quite clearly that there's room for improvement."

The researchers tested DREs from five companies, including Diebold, ES&S and Hart InterCivic.

In addition, voters seemed to approve of verification systems such as printouts that accompany some DREs, but the verification systems didn't significantly cut the error rate of DREs and often caused confusion and prompted voters to seek help from poll workers, said the study, conducted by political science and computer science professors from the University of Maryland, the University of Rochester and the University of Michigan. The study has been published in a book, "Voting Technology: The Not-so-Simple Act of Casting a Ballot."

Some of the study's results were surprising, said co-author Richard Niemi, a political science professor at the University of Rochester in New York. Niemi expected the volunteers would find paper ballots easiest to use because of the familiarity, he said. The top-rated DREs came out ahead of paper when voters were asked about ease of use and confidence that their ballots would be recorded as cast, he said.

"I certainly expected ... that the paper ballot would be the standard by which everything else would be compared," Niemi said.

The researchers, when testing vote verification methods, tested multiple systems, including printouts showing the voters' choices, a separate monitor that accompanied a DRE, and a receipt-type system in which voters could later check online or on the phone to see if their ballots had been accepted. Most verification systems increased the vote accuracy only slightly and caused confusion, with 5 percent to 8 percent of voters needing help, said Michael Hanmer, a political science professor at the University of Maryland.