The figures are from a study by electionline.org, a nonpartisan clearinghouse for election reform information. It is the most comprehensive look yet at how states implemented the major change to grow out of the 2000 presidential vote in Florida, when administrative errors and voter registration database problems kept thousands of eligible voters from casting ballots.

In the 43 states where data were available, provisional votes accounted for just over 1 percent of the total votes counted. In Alaska, 7.2 percent of all the votes counted came from provisional ballots, the highest of any state.

A law enacted by Congress in 2002 required all states to adopt procedures to allow people whose names are not on voter lists but who believe they are registered to cast ballots that can be checked later to verify their eligibility. The law left it up to states to implement the provisional voting rules.

Provisional voting "was a success in many ways in terms of what happened in 2000 when people were turned away and had no fail-safe way of voting," said Elizabeth Schneider, one of the authors of the study. "With provisional balloting in place, a majority of the people were not turned away; they were given a chance to vote."

The electionline.org study found that 70 percent of provisional ballots were counted in states with rules that deemed such ballots valid if cast anywhere in the voter's county or township. In states that required provisional ballots to be cast in the voter's correct precinct to be valid, only 62 percent were counted.

Counting procedures also varied from county to county in some states. In Arizona, for example, state rules required that provisional ballots must be cast in the correct precinct to be counted. But at least two counties, Gila and Pinal, counted provisional ballots that were cast in the wrong precinct.

While Alaska counted the highest percentage of provisional votes, five other states -- Oregon, Washington, Nebraska, Ohio and Colorado -- counted more than 75 percent. Five states besides Delaware -- Hawaii, Oklahoma, South Dakota, Kentucky and Indiana -- counted less than 15 percent of such ballots.

The study cautioned that comparisons between states are difficult because of widely varying rules governing such ballots. In Vermont, for example, only 101 provisional votes were cast and just 37 counted. Unlike other states, Vermont allows voters to make a sworn affidavit at the polling place on Election Day and cast a regular ballot.

The most significant difference among state rules for counting provisional ballots is the requirement over where they must be cast. Twenty-nine states count such ballots only if they are cast in the voter's correct precinct, while 17 count them if they are in the correct county or township. Four states are not required to offer provisional ballots because they permit voters to register on Election Day, and one -- North Dakota -- does not require registration.

The electionline.org study included data from 43 states and the District of Columbia. Missing were Mississippi, New York and New Jersey, which have not reported the number of provisional ballots cast and counted, and the five states that do not have provisional voting.