COLLEGE PARK -- University of Maryland Professors Paul Herrnson and Benjamin Bederson know as much as almost anyone in the country about voting machines.
And to do their job well, they need to try to forget all of it.
That's because the pair is part of a growing group of researchers who are undertaking the long-neglected work of studying voting technology -- determining which machines are the most reliable and easy to use, and how they can be improved.
To do this, the researchers are delving into the intricacies of new technologies and ballot designs. More importantly, though, they are putting themselves in the place of the lowest-common-denominator voters -- those who lack experience with voting or computers and who might, if left confused at the polls, vote differently than they intended.
As the country found out two years ago in Florida, the researchers note, a few perplexed voters can have a huge impact.
"If you have a small percentage of people having problems, it could be just enough to decide an election," said Herrnson, a political scientist and director of UM's Center for American Politics and Citizenship.
Herrnson and Bederson recently released the results of a survey of more than 1,200 voters on Election Day last month in Prince George's and Montgomery counties, which introduced new touch-screen voting machines this year. About 3 percent of the voters reported mechanical problems, and 7 percent said the machines, which will likely be adopted across the state, were not easy to use.
The pair's work is extending well beyond the state's borders. The Florida debacle -- in which a particular style of ballot confused many voters, while many punch card ballots failed to register clearly -- has fueled a small boom industry in voting research, and the two professors are in the middle of the action.
Last school year, Herrnson and Bederson won a $35,000 grant from the National Science Foundation to lay out the groundwork, with researchers at other schools, for a comprehensive study of voting technology.
The award was just a blip in the $350 million in research funding collected by the university last year, double what UM received five years ago. But the award could have big consequences: Herrnson and Bederson are in the running for a $1.5 million NSF grant to study how voting machines could be improved.
Studies by other researchers are under way as part of the $3.9 billion election reform bill passed last month by Congress. The vacuum of information about voting procedures is fast being filled, officials say.
"There is going to be a lot more money for experts to take a look," said Penelope Bonsall, director of election administration at the Federal Election Commission.
Herrnson and Bederson are focusing on how voters interact with new ATM-like touch-screen machines and optical-scanning machines that are -- thanks to a burst of funding since the Florida mess -- rapidly replacing punch cards, lever machines and paper ballots.
While the new machines are in many ways a vast improvement, the two researchers have found they aren't perfect. The touch-screen machines are especially in need of fine-tuning, they say.
It's hard to believe, they argue, but there are still people who have little experience with computers or ATM machines. These voters need the clearest possible instructions for navigating touch-screen voting.
For instance, some voters don't realize that to change their vote, they need to re-press the name of the candidate they don't want before they can press the name of the one they want.
Others accidentally press the screen to submit their vote before they're ready to, because that item is tucked at the bottom of the screen where voters review their choices -- a problem polling stations warned voters about last month.
"That's Design 101: If you have to give instructions on how to use this, there's something wrong with the interface," said Bederson, a computer scientist and director of UM's Human-Computer Interaction Lab.
In the survey results, some Maryland voters reported having trouble sticking their voting card into the machines to activate them, while others had trouble picking a language preference. About 17 percent of voters said they had help from poll workers.
The researchers warn of a broader drawback as well: Touch-screen machines don't leave behind a paper record of votes, in case the computer count should fail. Optical-scanning machines do give a hard copy: Voters fill out a ballot by hand, then scan it into the machine, which asks them to confirm their votes; the ballots are then kept as a backup.
Defenders of the touch-screen machines note that lever machines also don't leave a paper record. And they point out that machine manufacturers are finally allowing outside review of their designs, to assure that the programming is sound.
"We're confident in the testing this equipment has undergone," said Linda H. Lamone, Maryland's elections administrator.
This doesn't satisfy the UM researchers, who argue for upgrading touch-screen machines so they spit out a paper receipt with each voter's tally, just as a bank machine does. Lever machines, they say, could only be doctored one at a time, while the programs in computer machines could conceivably be flawed on a grand scale.
"If there is a serious problem, how do we go back and understand what happened? We can't, because it's entirely electronic, with no external, auditable trail," Bederson said.
The researchers plan to take a year testing machines in laboratories and in public places such as shopping malls. Ideally, they say, adjustments can be made for the 2004 elections, when many will encounter the new machines for the first time.
The stakes are high, they say.
"If you don't have the complete trust of voters, democracy falls apart," Bederson said. "And already, a significant number of people are losing trust."