PARIS -- Scientists who falsify, fabricate or plagiarize data are probably far more numerous than those caught out, reports a study to be published on Thursday.

Nearly nine percent of 2,012 researchers surveyed at 605 institutions in the United States claimed they had witnessed potential scholarly misconduct in the preceding three years, according to the study.

The 192 scientists who said they had spotted cheating colleagues reported a total of 265 separate incidents, many documented by hard evidence of ethical wrongdoing.

But less than 60 percent of these alleged acts were reported to the institutions concerned, and an even smaller percentage were in turn referred to the Office of Research Integrity (ORI) in the US government's Department of Health and Human Services.

The number of investigations submitted to the ORI, which conducted the survey, only averages 24 per year.

"Extrapolating the survey results -- even conservatively -- projects an alarming picture of under-reporting," concluded the investigators in a commentary, published in the British journal Nature.

Led by Sandra Titus, the ORI team calculated that at least 2,325 possible cases of scholarly misconduct are observed by colleagues every year among the approximately 155,000 researchers who receive financial support from the National Institutes of Health (NIH) annually.

If only 58 percent of these incidents are reported to the universities and institutes where these scientists work, "almost 1,000 could be assumed to go unreported to any official," they note.

There are many reasons why individual scientists might hesitate to report wrongdoing: doubts about what they have seen, fear of retribution, sympathy for colleagues.

Universities and research institutes made aware of cheating do not always carry through with investigations either, for fear of causing scandal or losing funding, suggests the study.

The commentary calls for more rigorous standards.

"Institutions must establish the culture that promotes safeguards for whistleblowers and establish zero tolerance both for those who commit misconduct and for those who turn a blind eye to it," the investigators said.

One of the more spectacular recent cases of academic fraud occurred in 2005, when the prestigious US journal Science was forced to retract an article it had published in which South Korean scientist Hwang Woo-Suk claimed a major breakthrough involving human stem cells.

A probe by Seoul National University found that the experiments had been faked.