Minnesota Public Radio News Producer Sanden Totten agreed to subject himself to an experiment. He took two reams of paper -- there are supposed to be 500 sheets in a ream -- and he counted them by hand. One by one.
"Oh boy," Sanden said as the 4-pound stack of paper crashed down on the desk in front of him. "This could take awhile."
And with that, he started counting.
Don't feel too bad for him. Election officials around the state collectively counted to 2.9 million during the recount. And they did it with lawyers, campaign volunteers and the media staring over their shoulders.
The amazing thing is, in more than half of Minnesota's 4,000-plus precincts, those human beings came up with exactly the same vote totals during the recount as Minnesota's vote-counting machines found on election night.
Secretary of State Mark Ritchie said the recount proves, once again, Minnesota's voting equipment is first rate.
"Incredibly accurate, these machines," Ritchie said.
But the machines and the people don't always agree. Throughout the recount process, both campaigns had representatives on hand. They challenged thousands of ballots in an effort to reduce each other's vote totals.
So, in precincts with challenged ballots, the picture is still murky. We won't have final vote totals for those precincts until the State Canvassing Board rules on the challenges.
But, even in precincts without challenges, the machine totals and hand counts didn't always match. They'd be off by a vote or two in one direction or another, sometimes more. Not counting the precincts with challenged ballots, there was a nearly 700-vote discrepancy between the machine count and the human count. That's a difference of less than one-tenth of 1 percent.
Ramsey County Elections Manager Joe Mansky said even though the hand counts in some precincts differed from the election night totals, we shouldn't blame the machines.
"It's actually not the ballot counter that is at error," Mansky said. "It's the people using it."
Sometimes election judges make mistakes. They forget to feed some of the ballots into the machines or they feed some of them through twice. But Mansky said most of the errors happen because voters don't follow directions.
Most Minnesotans vote by filling in an oval next to a candidate's name. If you do that, make your mark nice and dark. Mansky said the machine will correctly count your vote 100 percent of the time. But, if you make a faint check mark or circle the candidate's name, the machine isn't going to register your vote.
Mansky said humans are much better than machines when it comes to determining whom a voter intended to support. But, he said machines are far superior when it comes to actually counting up the votes.
"At some point people start to get tired," he said.
Exhaustion set in for Sanden around the late 400's.
"487, 488, 489, 450," he counted aloud confidently.
Luckily he caught his mistake.
"Generally speaking, our baseline error rate for hand counting paper ballots is about five per 1,000," Mansky said. "Our optical scan equipment is accurate to about two per 1,000."
In most elections, that's an acceptable error rate. But the margin in Minnesota's U.S. Senate race is so tiny, it makes a difference. All those little human errors add up and the 700-vote discrepancy the recount already picked up is more than three times the number of votes separating Norm Coleman and Al Franken.
Ramsey County's Joe Mansky said one of the good things about the recount is it gets people to pay attention to our voting systems. He predicts voters will be more careful next Election Day when they're filling in their ovals. Election judges will be more careful when it comes to operating the machines and he expects the Legislature will look at making some changes to the system too.
As for Sanden, he counted 1,016 sheets of paper in the two reams. There were only supposed to be 1,000.
So who was right, man or machine? Sanden or the factory that packaged up the paper? It's hard to say. All we can really do is have him count them again.