The nation's premier database of biomedical research contains as many as 200,000 questionable publications, ranging from possible outright plagiarism to researchers recycling their own work, Dallas scientists have found.

A team at UT Southwestern Medical Center came up with that number after using a computer program to scan summaries of nearly half of 17 million scientific and medical publications in the database. Originally developed as a way to search for information in the massive biomedical literature, the computer program has now resulted in investigations of possible scientific misconduct. The researchers say that between 1 percent and 2 percent of all publications in the database include duplicated text.

"I think this has a high probability of being a very good deterrent in the future," said Harold "Skip" Garner, the professor of biochemistry and internal medicine at UT Southwestern who led the study, "and ultimately improving the quality of all of the scientific literature."

A commentary on the computer program, written by Dr. Garner and colleague Mounir Errami of UT Southwestern, appears in today's issue of the journal Nature.

The word crunching from the UT Southwestern computer analysis highlights for scientists what most high school and college professors already grapple with – in the Internet age, copying and pasting is easy and can go unnoticed. But until now, Dr. Garner and Dr. Errami said, there was no efficient way to scrutinize the biomedical literature. Duplications were noticed primarily when researchers or journal editors read something they remembered reading before.

So far, UT Southwestern scientists have inspected 75 instances of duplications by different authors. Some Dallas researchers may be victims of plagiarism, the database indicates.

Dr. Joseph Kuhn, director of general surgical research at Baylor University Medical Center at Dallas, was among those whose work appears to have been substantially copied. UT Southwestern researchers reviewed a 2002 article by Dr. Kuhn and colleagues and found that in 2007, researchers from France published a highly similar article in a different journal. Overall, 40 percent of Dr. Kuhn's article was duplicated in the French article, including the entire introduction and methods section, the UT Southwestern database indicates.

Dr. Kuhn said he didn't know about the French paper.

"I'm not terribly surprised," he said. "The temptation to copy and paste is not just in high schools."

Dr. Roy Fleischmann, a rheumatologist and clinical professor of medicine at UT Southwestern, may also be a victim.

About 55 percent of a 2003 article by Dr. Fleischmann is duplicated in a 2004 article in a different journal, authored by Dr. Lee Simon, a rheumatologist at Beth Israel Deaconess Medical Center and Harvard Medical School in Boston.

Dr. Fleischmann said he was upset when notified of the duplication by Dr. Garner but is reserving judgment until the case is resolved by the two journals.

"It's distressing," he said. "You spend three or four months writing an article, and then it's duplicated. I don't know whether the second journal asked for permission to duplicate."

Dr. Simon, speaking through a representative, declined to comment and directed questions to the journal in which he published. The journal's editor did not reply to an e-mail inquiry.

Dr. Garner and Dr. Errami, along with other researchers at UT Southwestern, used the computer program to search 7 million entries in Medline, the National Library of Medicine's online catalog of scientific and medical journals from around the world. The duplications uncovered in the search have been deposited in an online database called Dejavu.

The scientists turned up two main categories of duplications. The first were instances in which similar text showed up either months or years apart, from different sets of authors. One obvious explanation for these cases would be plagiarism, Dr. Garner said. The other category included cases in which the same authors published highly similar text at different times, and sometimes in different journals.

Both categories include questionable scientific practices as well as possible violations of the original journal's copyright. However, Dr. Errami said, there are legitimate reasons for some of the duplications, such as study updates or simple cataloging errors in Medline.

"Some of them will be valid ... but it won't be the majority," Dr. Errami said.

The computer program compares only the abstracts, or summaries, that appear at the beginning of almost every scientific or medical research article. To see what's behind any instance of duplication, Dr. Errami said, you have to look at the full text of the article.

Duplication of another's work is an obvious no-no in the academic world. But duplicating one's own work is more of a gray area, said Raymond de Vries, a sociologist with the bioethics program at the University of Michigan in Ann Arbor. In recent years, tight funding has put scientists under extra pressure to publish, he said, and that often means publishing new articles that are only slightly different from each other.

"Scientists are saying, 'I want to let my data out bit by bit by bit,' " Dr. de Vries said.

Still, merely recycling one's own work to stretch a résumé hurts the scientific enterprise, said Thomas Mayo, a lawyer and medical ethicist at Southern Methodist University.

"You have an unethical gaming of the system by an individual who gets two or more credits from one article," he said. "That's the measure for promotion, salary, grant awards. If you have rigged those numbers, the rewards that come your way are a form of ill-gotten gain."

Dr. Mayo said he thought universities might want to use the UT Southwestern computer program when hiring new faculty or considering faculty for promotion.

Dr. Alfred Gilman, Nobel laureate and dean of UT Southwestern Medical School, said he thought journals would use the computer program. But for universities, he said, such searches might be more cost and effort than they are worth.

"There is some level of bad behavior, and that is unfortunate," he said. "But I think it's pretty low. It would be an awful lot of work for very small gain."