MIAMI - Even after all of Florida's disputed ballots from last fall's presidential election were examined by hand - every dimpled and hanging chad, every ballot containing more than one vote for president - the contest for the state's 25 electoral votes and the White House is a split decision.
Neither Democrat Al Gore nor Republican George W. Bush amassed a commanding, unambiguous lead. The outcome depends on how you count the ballots.
The Miami Herald, Knight Ridder Newspapers, USA Today and several Florida newspapers reviewed more than 176,000 ballots that were rejected by counting machines Nov. 8.
This review included the first statewide examination of more than 111,000 overvotes, ballots not counted on Election Day because they appeared to contain votes for more than one candidate.
Four different standards can be used to count the punch-card ballots that didn't record any votes for president when they were run through the counting machines, and the review examined ballots under all four.
Bush would have prevailed under the two most restrictive. His biggest margin would have been 407 votes under the standard most commonly accepted by states that use punch-card ballots. It requires that two corners of a ballot's chad be detached for the vote to count.
Gore would have won under the two most permissive standards. His biggest margin would have been 332 votes if dimpled chads, which bulge out but are attached at all four corners, were considered valid votes.
Those margins represent about 0.006 percent of the 5.9 million votes cast by Floridians.
The 2000 presidential election was the closest in 124 years, and Bush claimed the White House only after he was assured of Florida's electoral votes.
But because Gore won more popular votes and because the Florida results were so controversial, a number of people continue to question the legitimacy of Bush's presidency.
After an exhaustive hand recount of every disputed Florida ballot, it's clear that the answer to the question of who really won Florida will always be inconclusive.
Early last month, the Miami Herald-Knight Ridder-USA Today vote-counting project reported that if a statewide recount ordered by the Florida Supreme Court had been allowed to proceed, Bush almost certainly would have been declared the victor.
That newspaper recount project dealt only with 64,248 so-called undervotes, ballots that didn't register a choice for president when counted by machine.
This time, the newspapers tried to answer a broader question: Who would have won if all the disputed ballots were examined - all the undervotes plus the 111,261 so-called overvotes.
When the overvotes were reviewed by hand, 97 percent of them had to be thrown out because they contained votes for two or more presidential candidates. The rest, more than 3,000 ballots, could be awarded to a candidate because voters had selected a candidate twice.
For example, some ballots marked for Bush or Gore also had the same candidate's name written on them. Those ballots were rejected by the counting machines. Gore gained 682 votes on Bush from these ballots.
The study strongly suggests that of the Floridians who made mistakes that invalidated their ballots, far more intended to vote for Gore than for Bush. Three out of four ballots with more than one choice for president contained a mark for Gore. Only one of three such ballots had a mark for Bush.
"Gore would likely have won if all overpunched ballots had been properly marked, based on measures of voter intent," said Anthony Salvanto, a political scientist at the University of California-Irvine who analyzed the overvote results.
Those ballots probably would not have been reviewed unless one of the two candidates had asked for a statewide hand recount of every rejected ballot.
Gore considered doing that but decided against it. He requested only a recount of the undervotes in four counties where nearly one-third of all of Florida's votes were cast. Gore's campaign said at the time that he expected to get enough Democratic votes in those counties - Miami-Dade, Broward, Palm Beach and Volusia - to win the state and the White House.
The ballot reviews underscore what election officials have known for decades - elections are sloppy.
Many people don't know how to mark their ballots. The people or machines counting the ballots make mistakes. The precise numbers reported on election night are almost certainly wrong. None of those factors matters, however, unless an election is very close.
"It's amazing to me how some people misinterpreted how to vote," said Ronald Legendre, the chairman of Osceola County's canvassing board.