According to a survey conducted by Nature, some 70 percent of researchers have been unable to reproduce other researchers' experiments and about half haven't been able to reproduce some of their own.
More than 1,500 researchers took Nature's online about reproducibility in research and were asked whether they thought there was a reproducibility crisis, what might be causing it, and what could help improve research reproducibility.
Slightly more than half — 52 percent — of respondents said there was a significant reproducibility crisis and 38 percent said there was a slight crisis, Nature reports. Despite that, Nature says its respondents were hopeful as 73 percent said that more than half the papers in their field could be trusted, even though some studies showed reproducibility rates of 40 percent in psychology and 10 percent in cancer biology.
Respondents traced reproducibility issues to the pressure to publish, selective reporting, poor oversight, and more. Judith Kimble from University of Wisconsin-Madison notes that those factors are compounded by increased competition for grants and bureaucratic overhead that take time away from the lab. "Everyone is stretched thinner these days," she tells Nature.
Still, respondents said that improving experimental design, relying on better statistical approaches, and having better mentorship could all improve research reproducibility.
Pre-registration of studies is often touted as a way to combat selective reports, though Nature notes that only a dozen or so respondents mentioned it.
Though it's time consuming, Hanne Watkins from the University of Melbourne says that it's helped her work. "If it's built in right from the start," she adds, "it's just part of the routine of doing a study."