With paper ballots from the 2004 presidential election in Ohio scheduled to be destroyed next week, the secretary of state in Columbus, under pressure from critics, said yesterday that he would move to delay the destruction at least for several months.

Since the election, questions have been raised about how votes were tallied in Ohio, a battleground state that helped deliver the election to President Bush over Senator John Kerry.

The critics, including an independent candidate for governor and a team of statisticians and lawyers, say preliminary results from their ballot inspections show signs of more widespread irregularities than previously known.

The critics say the ballots should be saved pending an investigation. They also say the secretary of state’s proposal to delay the destruction does not go far enough, and they intend to sue to preserve the ballots.

“This is not about Mr. Kerry or Mr. Bush or who should be president,’’ said Bill Goodman, legal director of the Center for Constitutional Rights, a New York group that is part of the lawsuit. “This is about figuring out what is not working in our election system and ensuring that every cast vote counts.

“There is a gap between the numbers provided in the local level records, which until recently no one has been allowed to see, and the official final tallies that were publicly released after this election, and we want to figure out why that gap is there.”

After eight months inspecting 35,000 ballots from 75 rural and urban precincts, the critics say that they have found many with signs of tampering and that in some precincts the number of voters differs significantly from the certified results.

In Miami County, in southwestern Ohio, official tallies in one precinct recorded about 550 votes. Ballots and signature books indicated that 450 people voted.

Steven Rosenfeld, a freelance reporter formerly with National Public Radio, said the investigative team analyzed three types of sources. They are poll books used by officials to record the names of voters casting ballots, signature books signed by voters and used to verify that signatures match registration records, and optical scan and punch card ballots, used by 85 percent of the voters in the state. The rest used touch-screen machines.

“We’re not claiming that what we found reveals a huge conspiracy,” Mr. Rosenfeld said. “What we’re claiming is that what we found at least reveals extremely shoddy handling of ballots, and there are some initial indications of local-level ballot stuffing.”

In Miami County, Mr. Rosenfeld said, the team found discrepancies of 5 percent or more in some precincts between the people in the signature books and the certified results.

In 10 southwestern counties, he said, the team found thousands of punch card ballots that lacked codes identifying the precinct where the ballot was cast. The codes are typically necessary for the machines processing the ballots to “know’’ to record which candidate receives the votes.