Scientists can prevent the canonization –- general acceptance as fact -- of incorrect research results by publishing more reports of negative or inconclusive research.

That was the conclusion that Carl Bergstrom, biology professor at the University of Washington, reached after building a mathematical model of publication bias, defined as the tendency of journals to publish mostly positive research results.

Bergstrom pursued this research, published in eLife, out of concern that false claims could be canonized incorrectly, leading to an array of problems that can affect everyday life. He cites the example of the cause of stomach ulcers to illustrate the point.

Biologists originally postulated that bacteria caused ulcers. However, in the 1950s, gastroenterologist E.D. Palmer reported that bacteria could not survive in the human digestive tract. Instead, he blamed stomach acid. Fifty years later, using a better experimental model, researchers identified H. pylori as the culprit. Palmer’s work was falsely canonized.

Bergstrom’s work focused not on topics that attract a lot of research, particularly research results that yield identical results each time the experiment is repeated. Instead, they looked at less extensively studied topics. His team built a mathematical model that included rates of experimental error, how much evidence is needed to canonize a research result, and the frequency of publication of negative research results.

Researchers typically are under pressure to publish only positive results. The scientific community consequently lacks access to many failures or equivocal results that might have preceded, or followed, the positive results. Bergstrom’s model demonstrated that a lower publication rate for negative results led to higher risk for false canonization.

Raising the bar for canonization in the model did not reduce this risk. Increasing the percentage of negative results published did. The closer a piece of research is to canonization, the more impact negative results have.

"By more closely scrutinizing claims as they achieve broader acceptance, we could identify false claims and keep them from being canonized," said Bergstrom. He hopes these research results will encourage more exploration of bias from this and other sources, like funding sources, in scientific institutions.