Scientific research may be in decline across the globe because of growing pressures to report only positive results, new analysis suggests.
A study by the University of Edinburgh examined more than 4,600 scientific research papers published between 1990 and 2007 and found a steady decline in studies in which the findings contradicted scientific hypotheses.
Papers reporting null or negative findings are in principle as useful as positive ones, but they attract fewer readers and citations, so scientific journals tend to reject them.
It is acknowledged among scientists that this problem might be worsening, because competition in science is growing and jobs and grants are given to scientists who publish frequently in high-ranking journals. Many researchers, therefore, have speculated that scientists will increasingly pursue predictable outcomes and produce positive results through re-interpretation, selection or even manipulation of data.
The study examined research papers in which a hypothesis had been tested, in various scientific disciplines. Over the period studied, positive results grew from around 70 per cent in 1990 to 86 per cent in 2007. The growth was strongest in economics, business, clinical medicine, psychology, psychiatry, pharmacology and molecular biology.
The findings, published in Scientometrics, also show that papers reporting positive results are more frequent in the US than in Europe.
Dr Daniele Fanelli of the University's Institute for the Study of Science, Technology and Innovation, who led the study, said: "Either journals are rejecting more negative results, or scientists are producing more positives. It is most likely a combination of both.
"Without negative evidence in the literature, scientists might misestimate the importance of phenomena and waste resources replicating failed studies. The higher frequency of US papers reporting positive findings may suggest that problems linked to competition are greater in the US than elsewhere."
Explore further: Super Bowl athletes are scientists at work